US3116410A - Simple general purpose digital computer - Google Patents
Simple general purpose digital computer Download PDFInfo
- Publication number
- US3116410A US3116410A US856183A US85618359A US3116410A US 3116410 A US3116410 A US 3116410A US 856183 A US856183 A US 856183A US 85618359 A US85618359 A US 85618359A US 3116410 A US3116410 A US 3116410A
- Authority
- US
- United States
- Prior art keywords
- word
- instruction
- bit
- command
- accumulator
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Lifetime
Links
- 230000004044 response Effects 0.000 claims description 3
- 238000003860 storage Methods 0.000 description 98
- 238000012546 transfer Methods 0.000 description 24
- 230000008859 change Effects 0.000 description 18
- 238000006073 displacement reaction Methods 0.000 description 17
- 210000004027 cell Anatomy 0.000 description 16
- 239000004020 conductor Substances 0.000 description 14
- 230000004087 circulation Effects 0.000 description 11
- 238000010586 diagram Methods 0.000 description 10
- 241001289435 Astragalus brachycalyx Species 0.000 description 9
- 235000002917 Fraxinus ornus Nutrition 0.000 description 9
- 230000006870 function Effects 0.000 description 9
- 238000000034 method Methods 0.000 description 6
- 230000008929 regeneration Effects 0.000 description 6
- 238000011069 regeneration method Methods 0.000 description 6
- 238000012545 processing Methods 0.000 description 4
- 238000012360 testing method Methods 0.000 description 4
- 238000011022 operating instruction Methods 0.000 description 3
- 230000008520 organization Effects 0.000 description 3
- 230000001360 synchronised effect Effects 0.000 description 3
- 230000009471 action Effects 0.000 description 2
- 230000000295 complement effect Effects 0.000 description 2
- 230000001143 conditioned effect Effects 0.000 description 2
- 230000003111 delayed effect Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000002156 mixing Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 230000003068 static effect Effects 0.000 description 2
- 210000000352 storage cell Anatomy 0.000 description 2
- 241000272168 Laridae Species 0.000 description 1
- 241000275031 Nica Species 0.000 description 1
- 230000003466 anti-cipated effect Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000002457 bidirectional effect Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000003780 insertion Methods 0.000 description 1
- 230000037431 insertion Effects 0.000 description 1
- 230000005415 magnetization Effects 0.000 description 1
- 230000014759 maintenance of location Effects 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 208000003580 polydactyly Diseases 0.000 description 1
- 230000003134 recirculating effect Effects 0.000 description 1
- 230000001172 regenerating effect Effects 0.000 description 1
- 238000012163 sequencing technique Methods 0.000 description 1
- 230000007704 transition Effects 0.000 description 1
- 230000001960 triggered effect Effects 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/30—Arrangements for executing machine instructions, e.g. instruction decode
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F15/00—Digital computers in general; Data processing equipment in general
- G06F15/76—Architectures of general purpose stored program computers
- G06F15/78—Architectures of general purpose stored program computers comprising a single central processing unit
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/30—Arrangements for executing machine instructions, e.g. instruction decode
- G06F9/30003—Arrangements for executing specific machine instructions
- G06F9/3005—Arrangements for executing specific machine instructions to perform operations for flow control
- G06F9/30069—Instruction skipping instructions, e.g. SKIP
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/30—Arrangements for executing machine instructions, e.g. instruction decode
- G06F9/30003—Arrangements for executing specific machine instructions
- G06F9/30072—Arrangements for executing specific machine instructions to perform conditional operations, e.g. using predicates or guards
Definitions
- This invention relates to general purpose serial cornputer apparatus. More particularly, in an important aspect, the invention relates to a digital computer having at least two synchronized information circulating loops.
- the relatively short or minor loop is synchronized to the relatively long or major loop in an epicyclic timing relation.
- the computer operating cycle is based upon the recirculation of the minor loop and is divided into two phases. During the first of these one of many command codes is extracted from the minor loop. During the second of these phases the extracted command is executed upon a serially presented operand also circulating in the minor loop and, in certain instances, upon a second operand concurrently presented in serial fashion by the major loop.
- This computer is characterized by a displacement type of relative addressing.
- this type of addressing the address of the group of commands next to be executed by the minor loop may be a predetermined distance removed frorn the location in the major loop of the group of commands last transferred to the minor loop.
- this address may be determined by specifying the amount of the displacement of this next group along the circulating major information loop relative to the current position of the minor loop.
- addresses are absent in certain operations involving transfers from the major to the minor loop.
- Our invention discloses a computer which is intermediate on the one hand between a Turing type of computer, which by operating in a bidirectional serial manner on one binary bit at a time, uses a minimum of equipment at the expense of very slow operation, and on the other hand, a large scale stored-program parallel type of computer, which achieves very high speed operation at the expense of voluminous and complex equipment. Because of its efciency, our computer is capable of performing calculations involving business and scientic general purpose programs economically at speeds compatible with electromechanical input and output devices. At the same time computers in accordance with our invention require far less equipment than other general purpose computers. In addition, our invention teaches a computer organization which permits the stored program corresponding to a given computational purpose or format to be injected into the computer by a separate program loading device which can be shared by many such computers with consequent economy.
- a basic principle of our invention is to use a minor circulating memory loop of reference information in such a Way that the minor loop at any instant contains both a group of operating instructions and multiple digit accumulator information.
- Our invention in one preferred embodiment, then divides the time of circulation of this minor loop into two phases or half cycles during the first of which, designated precess time, one of the instructi'o ⁇ ns of the group is shifted out of this minor loop into an instruction register. During the second phase, designated execute time, the remembered instruction may be executed. Due to this automatic precession of the in structions, each instruction of a relatively large group may be executed in sequence without need for further transfer of instructions from the major loop until the entire group of instructions in the short, or minor loop is exhausted.
- a computer in accordance with our invention may be designed to execute directly as micro-operations such operations as storage to accumulator and accumulator to storage transfers, subtract, add, various jumps and conditional operations, and the input and output of a single character.
- a micro-program instruction corresponds to the performance of a single such operation upon every bit of a following word. It a micro-program instruction is modified during an execute cycle, the modified instruction advantageously may be executed during the remainder of the cycle.
- structures in accordance with our invention employ time sharing of components made possible by the division of the circulation of the short circulating loop into the separate time phases of precessing and executing.
- the elements of our invention include the two foregoing memory loops each with a reading station and a writing station.
- the writing station is downstream from the reading station so that information at a given location may be processed and replaced with the result of the processing at the same location.
- the reading station is advantageously downstream from the writing station.
- the information in the short loop advantageously may be continuously regenerated.
- These loops are preferably in the physical form of tracks on the surface of a magnetic drum, but, within the scope of our invention, they may also comprise a magnetic tape, a shift register, or any type of self-regenerating delay line.
- These loops circulate information consisting of words each composed of a fixed number of binary bits. These equal length words may be either an instruction word, or a number word.
- information is stored in a channel having a xed number of bits.
- the information stored in such a channel is divided in'to a xed number of storage sectors, say seventy-nine such sectors.
- this number is an odd number as later appears.
- the information of such an illustrative sector is further divided into groups.
- each sector contains a single binary indicating bit, a flag bit, and a word having thirty-two additional binary bits.
- Each word is further subdivided into a plurality of subgroups which together include the remaining bits assigned to a sector. These subgroups may, for example, have informational significance such as the designation of arithmetic operations to be performed or, simply, the value of a number.
- TIMING Clock means are provided to signal bit time and end of word time, and these times are lnade to be simultaneous for both circulating loops. End of word time is chosen to be coincident with the last bit in each word, and is conveniently designated taz for the above example, where tu is designated ag bit time.
- a feature of our invention is the provision of an odd number of words or sectors in the general storage memory loop, and an even number of words (preferably two) in the short instructions-accumulator" loop.
- a revolution of the general storage loop is selected in which the desired new instruction word will correspond in time to the instruction part of the short loop, whereas when it is desired to nd a new operand (number word) from general storage, a revolution of the general storage loop is selected in which the desired number corresponds in time to the accumulator word in the short loop.
- a single stage binary counter (ip-flop) is provided to count the end of word time signals and thereby to provide the two logical signal levels dividing the computer time into precess time and execute time.
- this counter signals that precess time is completed, the proper gates are operated to decode the instruction left in the foregoing shift register and to execute this instruction on each of the bits of the following number word read from the instructions accumulator loop, or from the general storage loop, or both, depending on the command.
- a simple binary adder-subtractor having two inputs, one output, and a carry-borrow memory flip-flop which may be interrogated or reset by the logic of the system.
- this memory flip-flop serves the important function of being an independent, conditional reference llip-op for decisions; this is in contradistinction to the usual computer practice of using an accumulatoroverflow as the condition for a decision ⁇ Simultaneously with the execution of the order, the results of logical operation being performed may be recorded, bit-by-bit, either back on the instructions-accumulator loop, or on the general storage loop.
- all recording (writing) operations in our invention are of the type which automatically erase whatever previously recorded bits may have existed on the loop medium. Thus, a previously stored word is replaced with a new word.
- instruction signal combinations i.e. information subgroups for designation of operations to be performed, are of four bits length. These four binary bits give rise to sixteen available instructions which may, illustratively, be allocated between twelve arithmetic operations and four displacement addressing operations. These displacement operations are those in which, for special programming purposes, the normal application sequence of operating instructions to stored information is interrupted and a new set of operating instructions introduced.
- a skip instruction is represented by a combination of all zeroes. This skip instruction does nothing except to regenerate the number in the accumulator portion of the short loop. The importance of this instruction appears hereafter.
- the instruction is cleared from the shift regitser (replaced by binary zeroes) so that the process of order code precession is characterized by a discontinuous march of the low order Hag bit followed by these zeroes automatically injected from the shift register towards the high order end (last in time) of the instruction word.
- a simplifying feature of our invention is the choice of a binary one for the flag bit in combination with the rst three automatically following zero bits to comprise the four bit instruction for an automatic fixed length displacement command, designated jump.
- the displacement addressing of our invention is accomplished by what we designate a jump delay instruction.
- this instruction When this instruction is included in an instruction word, room must be left to include the binary displacement address at the low order end of the instruction. This binary address number thus replaces one or more tetrads of bits which would otherwise be instructions.
- the logical network On decoding the jump delay, the logical network registers this fact n a mode" flip-flip memory, and then does nothing for one execute period except to regenerate the information in the short loop. On the following precess period, thc normal instruction word precession (through the shift register) is changed and instead,
- the remainder of the instruction word which now comprises the binary number representing the displacement address, passes through one of the two inputs of the subtracting unit.
- the subtractors output is stored in the short instructions-accumulator loop. In this way, by properly setting the other input of the subtractor, the address number suffers a diminution by one unit each time the short loop undergoes one regenerative circulation, namely, each two word times. Meanwhile, the number word in the accumulator is being regenerated intact during the entire jump.
- Branching in our computer may be accomplished by a jump conditional order.
- the xed-displacement (jump order) seeking a new series of instructions is made to be conditional on the state of a condition indicator flip-flop, which can be made to be a memory for the success or failure or overflow of a previous operation. If this condition indicator memory fliptlop is not in the predetermined state to cause a jump, the logic is arranged to change the jump conditional command into a skip which regenerates the accumulator word in the short loop and proceeds to the next instruction in the current series.
- a jump conditional order followed by a jump delay order in which the delay or relative address is made to be equal to two circulations of the general storage loop, can be used to set up an iterative test or interrogation routine.
- the routine will continue until some external condition changes the state of the conditional reference flip-flop, thereby causing a jump and the starting of a new sequence of instructions.
- This interrogation routine is useful for entering input information if the computer must wait until it receives a keyboard signal, or in providing output information if the computer must also wait for a feedback signal from the printer signifying that it is no longer busy and is ready to receive a new print command.
- Another feature of our invention is the leap cornmand which inhibits the word counter action at 132, thereby effectively interchanging preceiss time and execute time, and making the number word in the short loop an instruction word and vice versa. This order is useful in abridging the two circulation epicyclic correspondence period, and in using instruction words which have been built up as a number.
- the remaining elements of our invention comprise an input register, an o-utput register, a tetrad counter to distribute input ⁇ and output signals between parallel and serial presentation, a mode flip-Hop to distinguish between arithmetic and displacement type commands, and at least one delay flip-flop for giving early or anticipated access to the short circulating loops.
- the gating and reset logic of the system is thus arranged to control the information flow between the reading and writing heads of the two loops, the shift register, the add-subtract net, and the input and output registers in accordance with signals received from the clock means, word counter, shift register, conditional reference ipflop, tetrad counter, and mode ip-llop.
- the input and output devices are provided with switches to signal to the logic their availability.
- structures are provided for achieving related principal objects. These objects are to eliminate quiescent time in the utilization of computer elements thus to simplify, to render more economical, and to heighten ithe reliability of computer structures.
- lt is a further feature of the invention that there are provided structures having a series of micro-program instructions sequentially circulating in a loop with an accumulator number so that each micro-program instructions, in turn, is executed upon the accumulator number.
- a further feature lies in the provision of structure for the relative addressing of gro-ups of instructions in ⁇ accordance with both programmed and unprogrammed instruction signal combinations.
- Another feature of the invention lies in a circulating information loop in ⁇ which an operand may be placed for being subjected to continuing modification in accordance with a group of instructions sequentially circulating with the operand in the information loop.
- An additional feature is the provision of structures in which instructions ⁇ are alternately processed out of a working information loop into a register and then executed on an operand following in the same loop.
- a still further feature of the invention is the provision of structures for accomplishing fixed displacement addressing in response to predetermined instruction signal combinations which selectively initiate transfers of information from a relatively long loop after a predetermined number of cycles of a relatively short loop.
- An additional feature is the provision of logical branching paths upon the displacement of a relative address which is partially conditioned on the state of a memory device.
- Another feature of the invention is a momo-ry arrangement in which a specific word location in a major loop corresponds in time to la given word location in a mino-r loop only on ⁇ alternate circulations of the major loop.
- FIGURE l is a generalized illustration, partially in block tand partially in isometric view of a preferred oomputer according to the present invention
- FIGURE 2 is a representation of a portion of the memory channels of the computer of FIGURE 1, showing the time correlation between cells of the timing channel and the corresponding cells of the A-I register;
- FIGURE 3 consisting of FIGURES 3a and 3l, is a series of idealized information ow diagrams, each corresponding to a different micro-program operation;
- FIGURE 4 is a timing diagram illustrating the operation of an input and output counter
- FIGURE 5 is a more detailed diagram of the reading and writing circuits and the relay circuits of the computer of FIGURE l;
- FIGURE 6, consisting of FIGURES 6a through 6p, is a series of block diagrams showing in detail the various gating networks included within the logical network of FIGURE 1;
- FIGURE 7 consisting of FIGURES 7a and 7b, is a pair of flow diagrams of a calculator program written for the computer of FIGURE l;
- FIGURE 8 is a block diagram illustrating information ow in an illustrative structure in accordance with the invention.
- FIGURE 1 is a diagram of a computer according to the present invention.
- a magnetic drum 20 driven by a motor 2.1 has a magnetizable surface which is divided into four drum channels.
- a timing or clock channel 22 is recorded, containing 79 evenly spaced binary word sectors, each having 33 cells.
- Each cell of u channel is capable of storing a single bit of information represented by a predetermined state of magnetization within the cell.
- Each of the 79 sectors of the timing channel 22 corresponds to storages for 33 binary bits.
- each sector of the timing channel 32 cells contain binary representing bits to mark the rst 32 bits of each word and the last cell of each word stores a binary 1" representing spot to identify uniquely the last bit of each word, which is designated x32.
- the individual magnetic spots within the timing channel 22 are detected by a T, timing read head 24 ⁇ which generates electrical signals and applies them to a timing pulse generating unit 26, which is described in greater detail in connection with FIGURE a below.
- a second track on the drum 20 is used to create a two word recirculating memory loop which is designated the Accumulatonlnstruction channel 28.
- the Accumulator- Instruction or A-I channel 28 is used to store a rst binary word representing a number, and a second binary word representing a series of instructions corresponding to operations to be performed upon the number.
- the circulating two word memory loop or register mechanized by the A-I channel 28, uses only 64 storage cells on the drum corresponding to 64 cells of the timing channel 22.
- the additional two storage cells necessary to store two 33-bit groups are provided by two static storage elements, more fully described below.
- Associated with the A-I channel 28 is a writing transducer, the A-lw write head 39 and, separated from it in the direction of drum rotation by a distance equivalent to 64 cells is a read transducer, the A-Ir read head 32.
- the remaining two magnetic channels of the drum are used as information storage channels and are designated general storage 1 or GS1 34 and general storage 2 or G82 36.
- general storage channels each contain 79 binary sectors each of which contains a single ag bit though in other embodiments of the invention a different number of Hag bits may be advantageous.
- bits other than iiag bits make up a word which is further divided into subgroups.
- the subgroups may have an instruction significance or a number significance. In conventional partance they are instruction words or number words. For clarity of discussion, however, these are considered as "subgroups of the larger "vt/ord.
- Each of the general storage channels 34, 36 has an associated read transducer, a G51r read head 38 and a G52r read head 4i), respectively.
- Information is entered into the general storage channels by means of a pair of write transducers, a GSIw write head 42 and a G82W write head 44, respectively.
- the writing heads of the general storage channels are displaced from the reading heads in the direction of drum rotation by a distance equivalent to six word sectors plus one ccll or 199 cells of the timing channel 22.
- This arrangement permits information from the general storage channels to be read from a given sector, operated upon in the A-I register and returned into general storage, in the same sector replacing the former information by processed information. It is therefore a relatively simple matter to update" stored quantities.
- the extra cell between the read and write heads is necessitated by the fact that the general storage channel operating circuits include a reading staticizer which introduces a one-bit delay that must be accounted for.
- the words in general storage have, as a low order or least significant bit, a binary l representing signal which is designated a at bit. More specically, in this embodiment, cach sector of the general storage, except for one, has this binary "1 ag bit. However, in a predetermined word of one of the general storage channels, the least significant or flag bit position is occupied by a binary 0. It is this 0 flag bit that serves as an index pulse to synchronize the circulating loops of the computer, and control the START and FILL operations.
- a logical gating network 50 is shown only in block form in FIGURE l but is explained in greater detail in connection with FIGURE 6, consisting of FIGURES 6a through 6p, described below.
- the logical network 50 includes the read and write amplifiers and staticizers necessary for operation of the drum and its associated read and write transducers. Also included within the block, are all of the and and or gates, amplifiers, and inverters, which are necessary to execute the various logical functions during the operation of the computer'. These gate circuits receive signals from the drum transducers during certain time intervals, operate upon them, and provide signals to the drum transducers representing the result of a predetermined operation upon concurrently presented information signals. The predetermined operations correspond to the micro-program instructions presented by the A-I channels in a prior word time.
- Each flip-fiop may be a conventional bistable multi-vibrator, well known in the art, having two inputs and two outputs corresponding respectively to each of the stable states of the Hip-Hop.
- Each ilipflop receives control signals upon conductors from the logical gating network 50 for placing the flip-flop in either a set or l representing state or a zero or 0" representing state, and which are designated respectively, by the alpha-numeric designation of the flip-flop preceded by the letter S for set and Z for zero.
- the set output conductor has the alpha-numeric designation of the flip-flop and the zero output conductor has the alpha-numeric designation primed.
- An L2 flip-flop 52 is connected to the logic network and is normally used as a read staticizer for information coming from the A-Ir read head 32.
- An F2 ip-tlop 54 and an F3 flip-flop 56 are connected to the logical gating network 50 to be used, at different times, as staticizers for the information read from the general storage read heads and, during certain operations further described below, the F3 flip-flop 56 is connected in shift register fashion to receive the output of the L2 Hip-Hop 52.
- Four ip-ilops, F4 58, F5 60, F6 62, F7 64, are, during certain operations, connected together during the precess period as a four-bit shifting register and, as such, staticize four-bit instructions to control operation upon the succeeding accumulator number.
- An F1 Hip-Hop 66 is used primarily as a carry-borrow Hip-flop in arithmetic operations. In other operations of a conditional nature, and in input and output operations, as will be described below, the F1 ilip-op 66 serves as a supplementary memory'.
- Another Hip-hop, the FS flipilop 68 may be considered a two-counter which assumes alternate stable states in successive word intervals.
- the F8 flip-flop 68 can therefore be used as a basic reference, the state of which marks the precess and execute periods, and therefore identifies the signals appearing at the A-Ir read head 32 of the A-I register as number signals or instruction signals. In one state, F8, the signals appearing are bits of an instruction word while in the other F8 state, the signals representing bits of number words are presented.
- An F9 ilip-op 70 is another two-counter that remembers which of the two general storage tracks is in communication with the computer. Each time a track change is instructed, the state of the F9 ip-fiop is changed. The output of the F9 flip-Hop therefore can be used as a logical input for head switching as between the general storage tracks, F9" selecting GS1 and F9 selecting GSZ.
- a [p32 flip-op 72 is provided to identify uniquely the time interval corresponding to the last bit time of each word.
- the tp32 flip-flop 72 is in a rst stable state, w32', during the first 32 bits of each word, but during the last or 33rd bit, designated 132, the hip-flop assumes its other stable state, tp32.
- the tpgz flip-flop 72 can also be used to identify the lowest order or ag bit at time tu lof the next succeeding word.
- the recirculation path of the A-I register normally includes both the L2 flip-op 52 and the F3 ip-fiop 56 to total 66 bits of storage
- the L2 flip-flop 52 will be staticizing the t or flag bit of the next word.
- An input-output unit 80 is provided to enable the computer to communicate with the external world.
- a plurality of input keys (not shown) are provided to generate a combination of binary signals on tive input conductors and a sixth control signal on a separate conductor.
- the logical gating network 50 applies a seventh signal to the input devices on a seventh control signal conductor.
- the gating network 50 provides a combination of signals to the output device on iive output conductors and a sixth control signal on a separate conductor.
- the output device returns a control signal on a seventh conductor.
- a multi-conductor input cable 82 transmits the signals between an input portion 84 of the input-output unit 80 and the gating network S0.
- a multi-conductor output cable 86 transmits signals between the gating network 50 and an output portion l88 of the input-output unit 80.
- a pair of relays R1 90 and R2 92 are provided to facilitate the starting and stopping of the computer.
- the manner of interconnecting the relays is shown in greater detail in connection with FIGURE d below.
- Logic provided within the logical gating network 50 is responsive to signals from the relays to control the energization and de-energization of the electromechanical and electronic components associated with the computer.
- each sector is 33 bits in length having a flag bit in the lowest order or to position and, in the case of number 10 words, 32 binary digits.
- An instruction word contains a flag bit and up to eight, four-bit instructions. In the case of certain relative address operations, however, an instruction word has less than eight instructions, with the remaining group of digits representing a binary number corresponding to a relative address.
- the cells of the individual channels are arranged in vertical alignment representing concurrence in time only.
- the individual cells in a given channel may have any physical position with respect to the cell-s of a different channel so long as the read and write transducers are positioned to present signals synchronized in time.
- the read transducers are indicated as windows on their respective channels and, in the portion of the A-I channel, the two read staticizers, the L2 and F3 flip-flops are also shown as windows on the channel.
- each half cycle is one word time in length, alternate half cycles being designated precess and execute word times.
- precess word time four bits of an instruction word, representing a single instruction which corresponds to an operation to be subsequently performed, are extracted from the instruction word and are staticized in a four-bit register.
- the staticized bits control the disposition of the number presented by the A-I register.
- the A-I register may be considered as an accumulator register, storing a number to be operated upon.
- an instruction word consists only of the ag bit in the most significant bit position followed by 32 zeroes.
- an instruction word is arranged with the ⁇ most significant bit of the first instruction located in the last or t32 bit position. It is understood that the flow of information is such that the recorded information moves to the right, passing under a stationary read transducer A-Ir indicated by the properly designated Window at the right side of the drawing.
- the first instruction to be executed includes four bits, in, i12, im and 14, stored respectively in the t32, x31, t3() and f2s bit positions.
- the bits of the second instruction, i2, through i2. are stored in cells corresponding to t28 through 125.
- the bits of the last instruction, 1'81 through i8' are located in the cells t., through t1.
- an instruction word includes a delay type instruction which re-quires a jump to some relatively remote location in general storage
- the instruction requiring this transfer is followed by a binary number which represents a relative address of the word being sought. This binary number is located at the least significant end of the instruction word and occupies as many cells as are necessary to specify the relative address.
- a typical number word 104 contains a flag bit at the low order end ⁇ and a plurality of binary digits which may form the number.
- the most significant digit position, t32, of the number word may be interpreted as a sign bit at the programmers discretion.
- Eleven of the sixteen primary operations, set out in Table 1 above, may be considered arithmetic instructions in that they utilize the accumulator.
- These eleven consist o-f the following: transfer to general storage (TS), transfer to accumulator (TA), extract-leap (E/L), output (OUT), input (IN), subtract (SUB), add (ADD), clear (CL), right shift (RS), add conditional (ADDc), and left shift (LS).
- TS transfer to general storage
- TA transfer to accumulator
- E/L extract-leap
- OUT input
- IN subtract
- SUB add
- ADD clear
- RS right shift
- add conditional (ADDc) add conditional
- LS left shift
- transfer to general stor- :age command is the only instruction which records new information in general storage.
- live of these instructions namely, transfer to accumulator, extract, subtract, add, and add conditional, utilize the general storage read head as the source of the second operand.
- jump group namely, jump (I), change track (CT), jump conditional (Jc), and jump delay (Irl).
- I jump
- C change track
- Jc jump conditional
- Irl jump delay
- FIGURES 3a-3k are especially useful in correlating the time sharing functions played by these flip-flops during each instruction, and in following the information flow paths for these operations.
- the F1 Hip-flop 66 is the general purpose conditional memory which is used during the relevant arithmetic operations as the carry-borrow flip-flop for the addersubtractor unit. 1n general, F1 remains static during precess word times (except during the jump delay and change track commands) so that in effect it carries information forth from a previous execute word time to a subsequent one.
- conditional operations are conditioned on the information content of F1, remembered from the previous operation.
- conditional part of the add conditional, extract-leap, and jump conditional commands is whether or not they are to be executed, whereas in the jump delay and the change track commands, the condition of the Fl flip-flop is not a condition to the execution, but instead modulates how much displacement will take place as determined by the delay address numbers in the given instruction word.
- F1 is usually referred to as the carry flip-flop (designated K flip-flop) although it actually serves as a multi-function conditional register.
- tiip-tlop F2 54 serves as the general storage read flip-flop in those operations where a general storage operand is needed.
- F2 teams up with the F3 liip-liop S6 to serve as a tetrad counter to identify time intervals to to t3.
- F2 serves as a counter Hipflop to designate I@ intervals during execute word times.
- F2 plays an entirely different role. Here it serves as a mode flip-flop to designate whether the computer is undergoing a normal precess word time or if one of the four jump commands (jump, jump conditional, jump delay, or change track) is being performed.
- F3 rather than the F2 ip-op serves as the general storage read tiip-iiop when a new instruction word is being transferred from general storage to the instruction portion of the A-I channel during a jump type operation.
- F2 represents the general storage read flip-flop for number words (during execute word times)
- F3 represents the general storage flip-flop, when necessary.
- F3 represents a second A-l channel read flip-flop, in series with the L2 flip-flop. F3 serves this same role during execute word times for most of the commands.
- F3 in conjunction with F2 serves as a tetrad counter (from to to r3) as previously mentioned, but in general F3 represents an A-l channel read flip-flop whenever it is necessary to use the accumulator as an operand, regenerate the accumulator, or normally precess the instructions.
- the F4 to F7 flip-flops, 58, 60, 62, 64 represent the four flip-flops in the command shift register, under control of the F8 tiip-tiop 68.
- F8 may be considered as the Word time counter, in that its sole function in the computer is to count sector pulses at t32. As such, it serves as a fundamental reference flip-flop.
- the command shift register advances the instruction Word from the A-I channel read head into the A-I write head, as is best noted in FIGURE 3b.
- the high order tetrad of the instruction word remains locked in the command shift register as F8 ends and F8' (execute word time) begins.
- this tetrad is decoded from the command shift register as one of the sixteen primary mirco-program instructions.
- the F9 flip-flop 70 represents the change track flipfiop, and so controls which of the two general storage tracks is selected for reading or writing. This control function is best seen in connection with FIGURES 5a and 5b, described below.
- the only time that F9 changes is during a change track instruction (CT).
- the change track instruction is identical in all respects to the jump delay instruction, except that the F9 ⁇ lip-Hop is triggered to its complementary or opposite state as the instructions name implies.
- the change track command is really a member of the jump type (displacement address) instructions identical in all ways to the jump delay except for its one special duty.
- F9 serves one additional purpose of serving as a fth channel of output to the typewriter or printer thereby giving the programmer a choice of 32 output characters rather than the 16 provided by the four binary digits.
- FIGURES 6er-6p comprise a detailed schematic of the logical gating network 50 from which those skilled in the art can reproduce an operating model of our invention.
- the regeneration of the accumulator requires the information word to pass from the A-Ir head, 32, back to the A-IW head 30.
- signals pass from the A-Ir head 312- LS-SLZ [FIGURE 611], G8-G12-G17-F30-SF3 [FIG- URE 6bl, G53-G63-L1 [FIGURE 6n], G20-L7 [FIG- URE 6d], to A-Iw head 30.
- the A-Ir head 32 leads through a non-inverting amplifier 110 to generate a signal on conductor L8 and through the inverting amplifier 112 to generate a signal on conductor L8'.
- a l is read by A-Ir head 32, a relatively high or l representing signal will appear at clock time C1 on lead L8 while if a 0 is read by head 32 a relatively high or 1" representing signal will appear within bit time C1 on lead LS'.
- FIGURE 6h these leads are shown applied to control the setting of the L2 flipliop, 52.
- the signal path then oontinues through the network designated F30 to control the setting of the F3 tiip-op 56 one bit time later, thereby generating either F3 or F3'.
- Gate G55 in FIGURE 611 is dependent on gate G53 and on the foregoing F3 signal, together with A10, to continue through gate G63, inverting amplifier 114 to L1'.
- the signal L is applied in FIG- URE 6d to gate G20, the output of which generates a signal L7 by a rst inverter 116 and signal L7' by a second inverter 118.
- the path leads to the A--lW head 30 via gate G1121, one shot pulse generator 120 and non-inverting amplifier 122 in the case of an L7 or 1" representing signal to be recorded or via gate G122, one shot 124, and non-inverter 126 in the case of an L7 or 0 representing signal.
- the F5 diode in G8 and the F6' diode in G53 allow this path to operate. It should be noted that the GS- G12-Gl7 part of this path does not regenerate the low order flag bit (tu) of the accumulator.
- the reason for this is that the F30 net (FIGURE 6b) is timed one bit early because of its position in the A-I register circulation loop (this bit represents the ⁇ unit delay in the F3 ipflop).
- the F3 llip-op contains the most significant bit of the instruction tetra-d, which is a 0" at x32 of the previous F8.
- the low yorder bit of the accumulator number is in L2 :and the F30 net.
- no gate of FIG- URE 6b can transfer the contents of L2 to F3.
- the F3 flip-flop therefore copies its contents into the writing circuits, thereby resetting the IOW order bit position to 0, thus preventing the regeneration of the low order accumulator bit.
- TRANSFER TO GENERAL STORAGE (0001) (1) The K flip-flop and the number in the accumulator remain unchanged. The flag or low order bit of the accumulator is reset to i). T'he number in the accumulator is transferred to general storage thereby replacing the number previously stored there as is shown in the ow path of FIGURE 3C.
- this command is identical to the skip command.
- the previous description of the skip command serves equally well with respect to K, the regeneration of the accumulator, and the ⁇ loss of the low order accumulator bit.
- the accumulator regeneration path given for ⁇ the skip command is identical for the transfer to general storage command.
- the path from the A-Ir read head 32 to ZF3-G18-L10'- (G114 or C117) and A-Ir read head 32 to SF3-Gi19- L11-(G114 or G117) represent the flow of information signals being transferred from the accumulator to the general storage write circuits.
- the rst of the two paths represents that taken by the Os in the information word, ⁇ while the second represents the path of the ls.
- the source of the number in this transfer is a general storage read head (38, 40), which, as has been previously noted, is six sectors removed from the general storage write heads.
- the actual path of transfer from general storage to the accumulator is L12-(G1, C109, G2, G3)-G4G7-F20-SF2 [FIGURE 6u), G60-G63- L1' [FIGURE 611].
- G20-L7 [FIGURE 6d).
- the F5 diodes in G3 and implicit in G1 through the A5 diode [FIGURE 6p] and the L12 diode in G109 and G2 [FIG- 15 URE 6a] allow the path to operate through G4 [F1G URE 6a].
- the F2 ipflop 54 which represents the general storage read llipflop in this command as is seen in FIGURE 3d, is reset originally as we enter the command. This means that is transferred into the low order bit position of the acetirnulator at to, and consequently the low order bit previously there is lost (reset to 0).
- K and general storage remain unchanged.
- the low order bit of the accumulator is lost (ire-set to 0).
- the number in general storage is logically multiplied by the number in the accumulator and the result is placed in the accumulator (FIGURE 3g).
- the path of the general storage operand up to this gate is L12-(Gl, 6109, G2, G3)G4-G7F20-SF2 [FIGURE 6u), G53- GSS [FIGURE 6:11, while the path of the operand from the accumulator is LS-LZ [FIGURE 611], GS-GlZ-GIT- F30-SFI) [FIGURE 611], G55.
- the FS' diodes in G1, G3, and G8., and the L12 diode in G1109 and G2 allow these paths to operate.
- gate G? (FIGURE 6p) is provided as an input to and gate G75 which is enabled at t32.
- Gate G10? is responsive to signals from the F3, F4, F5, and F6 flip-flops which store the command that will be in the command register at to. If this tetrad is F3', F4', F5, F6, and also if F1 ⁇ is set to "0 (i.e. Fl), then a leap command exists and circuitry should be provided to inhibit the triggering of the F8 flip-flop. This is accomplished by making the operation of the "and"aded G75 16 contingent on the presence of any one of.
- the first case to be analyzed in the output command will be where the busy signal E10 is found to be present.
- the busy signal E10 is found to be present.
- FIGURE 4a represents the timing diagram for the output operation.
- the F2 and F3 flip-flops 54, 56 combine as a two stage tetrad counter as set out in Table 2, below.
- the F2 F3 final condition extends across the rest of the word time, but at the end of t3, the output command code is reset to a skip command as shown by the state of the F5 ip-op 60 in FIGURE 4a.
- the K flip-flop F1 is set to a 1 in order to serve future notice that the output command was successful.
- the output cornmand only lasts four bit times before changing to a skip command.
- the F3 Hip-flop 56 serves as part of the accumulator loop with the number word appearing at its output from t1 to x32.
- the output command We wish to print out only the low order tetrad of the accumulator, which ordinarily is available in the F3 flip-iiop 56 at t1 to t4.
- the F3 ip-op part of the tetrad counter For efficient minimization, however, we wish to use the F3 ip-op part of the tetrad counter for output selection.
- F3 From t5 to In, F3 truly represents the remainder ofthe information from the accumulator (passing into F3 from L2). At tit. however, F3 represents an inherited liif. from the last state. t3, of the tetrad count. Since it is indicated to cicar both the low order tetrad and the ting bit, it is necessary for the net Lt to produce G's from tanti. In order that the t4 bit be zcrocd, the tetrad count for I3 should include F3' rather than F3, and the tetrad counter therefore corresponds to the Gray or reflected binary code count.
- the path F-G'Y-Eln represents the fifth Wire in the five wire output code yielding thirty-two combinations. A more detailed description of this fifth Wire in the output code was previousiy described in conjunction with the primary fun-:tions of the F9 tiipdiop 7S.
- r1, t2, and t3 positions ot the accumui-.i if the keyboard character is a non-numeric charttiot one ol the digits through 9), a 1 is written e highes" order bit position (32) ot the accumulator. s icmai der ot the accumulator is cleared to all D's. IC the keyboard (1o-.vn signal Sift is not received, K is reset to 0 and thc accumulator fills with either zeroes or meaningless information (depending on whether ...f'ftches Si, S2, Sal, and are open or in the act ol bouncing or being set).
- Fl the K p
- FlGURE 4b shows a timing i .gram for both the successful and unsuccessful input.
- F11 is ally reset as the input operation begins. This :shed at 132 of the previous Fd by the path VFlGURE 6m] with the F3' diode making the path possible.
- Vl on the input timing diagram (FIGURE 4b) high it the ltey down signal SiO is received a successful input.
- the path is SIO-G2-G43- [FIGURE om i. Il the ltey down signal is not rcceivL Fl remains reset.
- the key down signal (515)) as received trom the keyboard is nothing more than a delayed common contact.
- the contacts providing the other input signals S1, S2, S4, SS, and S11
- the contact for the lzey down signal Sit itself may still n: bouncing. but this docs not matter since the system .s it only at one discrete time interval, namely I2 dicated. ..ier requirement oi the input timing system is that trop set F3 if Fl has been set at the cnd of t2.
- the path or" a 2 bit is SIZ-Gld-GIQLLAZ.
- the path of the 22 bit is Saf-*Glhlh-Glti- Y2.
- the path of the 23 bit during r3 is SSAGTSGESA.
- the path ior uic nonmuznric (not 0 through 9) indicator bit at tgz is biiGlll4Git5-A12.
- the low order tctrad of the accumulator is applied from i( to l( to the write head.
- the successful input operation records the low order tetrad in the accumulator in the t0 lo t3 bit positions, cle-.trs the accumulator (to s) from r11-tgl, and, il necessary, inserts the non-numeric tetrad indicator bit at tag.
- the clearing of the accumulator from t4 to rs1 is accomplished by simply writing no logic in the net (Ll) to do otherwise.
- a successful input command is usually followed by a shift left command, explained in greater detail below. This shifts the high order in bit (representing the nonnumeric tetrad indicator bit) into the K llip-ilop where the programmer can then test for its presence.
- the shift left command will also shift the low order tetrad from (t0 to t3), to (t1 to t4), and thus place it in its normal position in the accumulator.
- the input operation normally serves as the lirst of a group of four operations.
- the second operation of the group will be a jump conditional, explained below, which tests Whether or not the input was successful by checking the K Hip-flop. A shift left command (as before noted) then follows the jump conditional provided the input was successful.
- the fourth operation of this group will be another jump conditional which will test for the pres- -t ence of the non-numeric tetrad indicator bit by means of the K flip-Hop.
- the L5 gate circuit represents the sum-difference net for the adder-subtractor unit.
- the path of the resultant signal from L5 to the A-I write head is LS-GSS-G-Ll' [FIGURE 611], G20-L7 (FIGURE 6d].
- F1 represents the carry-borrow
- F2 represents the A major purpose of the K (F1) llip-ilop in our invention is to serve as the carry-borrow flip-flop for the adder subtractor unit. Consequently, at the end of an F8 executing an addition or subtraction operation, F1 will be left high (set to l) if wc have had an overflow in the addition or if we have subtracted a larger number from a smaller one. It should be observed that in general F1 does not change during an F8 word time, so that in essence it carries forward information from one F8' to the next F8. Thus if we followed a subtraction operation with a jump conditional (conditional on K1 to be explained below) we could branch the program depending on whether or not subtraction resulted in a change of sign.
- the three inputs to the addcr-subtractor unit of FiG. 6e are F1, F2, and F3.
- F1 represents the carry-borrow ilip-llop.
- F2 represents the operand from general storage in the addition and subtraction operation, and its path from general storage to F2 is LIZ-(Gl, G1G9, G2, GB)HG4G7"-FZO-SF2 (FIG. (im.
- the F6 diode in G1, the L12 diode in Gl@ and G2, and the F4 diode in G3 allow this path to operate.
- F3 represents the operand from the accumulator which is the augend and minuend in the addition and subtraction operations.
- the path of this operand from the A-I read head to F3 is LSLZ (FIG. 611], GS-GIZ-Gl-FSQLSFT (FIG. 611], with the FS diode in G8 allowing the path to operate.
- this mixing path in the adder-subtractor is simply to give a sum (or difference) on the L5 output line whenever any one or all three of the inputs (FI, FZ, and F3) are present.
- the sum-difference logic for L5 may be Written as set forth in Table 3B.
- the low order bit of the accumulator is reset (cleared to 0) in both thc addition and subtraction operations.
- F2 and F3 are low at tu simply because no logic was written to set them at x32 ot ⁇ the previous F8.
- the carry-borrow flip-flop F1 was reset to 0 at the end of F32 of the previous FS by means of the path G48-G49-G52-ZF1 (FIG. 6m] with the F3' diode allowing the path to opcrate.
- the path for setting F1 high as a borrow corresponding to Table 3B above is (F2, F3)-G82-A23 (FIG. 6Fl, G44G47 SFI (FIG. 6111].
- the corresponding reset path for F1 in subtraction is (FZ, F3)-Gl ⁇ -A2I (FIG. 63], G5l- The initial output 0 because the input GSZ-ZFl (FIG. 6rn].
- the path for setting F1 high as a carry is (FZ, F3)-Gtl-A22 ⁇ (FIG. 611], G42.-G43-G4i-SF1 IFIG.
- K is reset to 0 initially.
- the number in general storage is added to the number in the accumulator, and the result replaces the number in the accumulator. If the result in the accumulator overflows (exceeds thirty-two bits) K is set to a l.
- the number in general storage remains unchanged.
- the low order bit 0f l HCCUmUllO 1S cleared (reset to 0).
- JUMP (J) (1000) (1000)
- K. general storage. and accumulator remain unchanged.
- a new instruction word for general storage is brought into the instruction portion of the A-I channel during the next word time.
- the first (high order) tetrad of this new instruction word is executed three word times later (or equivalently four word times after the jump command was initially executed).
- the jump command represents the first of tour (jump, change track, jump condition, and jump delay) displacement addressing instructions. They are control instructions (in contrast to arithmetic instructions such as addition) in that they operate on the instruction portion rather than the accumulator portion of the A-l channel.
- the normal routine of setting up (precessing in) a new command during and then executing it during the next word time F3 can only be applied to the arithmetic type instructions and the skip command. This results from the fact that only the accumulator is available during Fti'.
- the control type instructions are set up during F8 as usual, but we find it impossible to execute them the next succeeding F3 word time, since the instruction portion of the channel is simply not available during F8.
- the purpose of the jump command is to transfer a new instruction word from general storage into the A-I channel.
- every instruction Word would only have seven useful commands, with the eighth command being used as a jump in order to replace the instruction word with a new orteA ln our invention, however. all eight commands in the instruction word are useful.
- the jump command need never be programmed as the eighth command in order to transfer in a new instruction word. The reason for this is the automatic jump feature.
- the automatic jump looks exactly as if a jump command had been programmed as a fietitious ninth command on the low order end of the instruction word.
- the automatic jump need not he programmed, the regular time ailotted for the execution of a jump command must also he allowed for the automatic jump when writing a program. ln other words, the programmer knows that the computer is going to introduce a ninth command, namely, the automatic jump, after all the instructions of a word have been executed. Thus we see that this computer has automatic sequential control for instruction words.
- the new instruction word When the new instruction word is transferred from general storage to the AHI channel (as a result of a jump type command), it occupies the position (L32 in the orders channel. During this transfer, however, a ls bit is always deposited into the low order (iiag) t0 position of the instruction portion of the A-l channel. The l ⁇ s bit will precess up the loop with the instruction word. It is this l ⁇ s bit that ultimately is executed as the automatic jump command.
- This preccssion is achieved during F8 by passing the A-I channel read output through the command shift register (four bit delay) and back to the A-Iw write head.
- G3t'!SF7 G3'i-G5ZE7
- the F2' diode in G10, and the tpgz' diode in Gll. and the Al diodes in G65, G27, G25, G35, G37, G30. G34, and G62 make this path possible.
- the F4 diode in gate G64 is redundant logically, but is necessary for signal loading considerations.
- the A1 signal in this path represents the "advance" signal for thc command shift register (F4-F7).
- the logic for this advance signal is represented by the path G93-G94-A1 (FIG. 6121.
- the path for the insertion of the l ⁇ s bit (representing the automatic jump) into the orders channel to t (when the instruction Word is being transferred from general storage to the order channel) is G74-A13 (FIG. Gpl, Gl5-Gl7-F3-SF3 (FIG. 6M, G81-A22 (FIG. 611], G56-G63-L1' IfFIG. Gull, G20-L7 (FIG. 6d'l.
- the ls bit is originally inserted (G15) at 132 of the previous F8', it picks up a unit delay in going through the F3 flipdlop so that it appears at the A-Iw write head (L7) at to of the following F8.
- the high order terad of the orders channel (which occupied the position rggg) is left in the command shift register, while the second highest order tetrad (tggga) is precessed up into the hielt order tetrad position.
- This tetrad which pre Waits oit the high order end of the A-l channel and left in the command shift register is the command that will be executed the next F3 word time.
- function net (which represents the common reset signal for the command shift register) resets the commend shift register to all G ⁇ s at the ends of tgt; or" every Fti'. This is achieved in the A2 net by the path AES-G92A2 snown in FIG. 6p.
- the four paths from A?. to the command shift register are AZ-G66-ZF4 (FIG. 60j, iXi-Gin UWG. tijl, AZ-GSS-ZF (FIG. (til. and .it2-Ga ⁇ lf-G3,3G35Z ⁇ 7 (FIG. 61'1.
- G3i inhibits the resetting of F7 during the change track and jump delay commands. The reason for this will be covered later during the description ot' these commands.
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Software Systems (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computer Hardware Design (AREA)
- Executing Machine-Instructions (AREA)
Description
Dec. 31, 1963 R. J. A MANNA ETAL 3,116,410
SIMPLE GENERAL PURPOSE DIGITAL COMPUTER 9 Sheets-Sheet 1 Filed Nov. 30. 1959 ram flamber- Dec. 31, 1963 R. J. L A MANNA ETAL 3,116,410
SIMPLE GENERAL PURPOSE mslm. COMPUTER Filed Nov. 30, 1959 9 Sheets-Sheet 2 n 42 i? 5 j a fz? I JZ 32 l fgj/ fx1/M- aud/f) f7!- Z (Mm/w@ i6 JZ .5? 5f #EH 52 55 VI-Ls Dec. 31, 1963 R. J. LA MANNA ETAL 3,116,410
SIMPLE: GENERAL PURPOSE DIGITAL COMPUTER 9 Sheets-Sheet 3 Filed Nov. 30, 1959 56 /rum l 71k) Dec. 31, 1963 R. J. LA MANNA ETAL 3,116,410
SIMPLE GENERAL PURPOSE DIGITAL COMPUTER Man @Parar/.n
Dec. 31, 1963 R. J. LA MANNA ETAL 3,116,410
SIMPLE GENERAL PURPOSE DIGITAL COMPUTER Filed NOV. 30, 1959 9 Sheets-Sheet 5 De@ 3l 1963 R. J. A MANNA ETAL 3,116,410
SIMPLE GENERAL PURPOSE DIGITAL COMPUTER Dec. 31, 1963 Filed Nov. 30. 1959 R. J. LA MANNA ETAL SIMPLE GENERAL PURPOSE DIGITAL COMPUTER 9 Sheets-Sheet 7 6 f.; d ,421 ,ad
fz an e De 31, 1963 R. J. LA MANNA ETAL 3,116,410
SIMPLE GENERAL PURPOSE DIGITAL COMPUTER ra @verf/aw Dec. 31, 1963 R. J. LA MANNA ETAL 3,116,410
SIMPLE GENERAL PURPOSE DIGITAL COMPUTER PiIed Nov. zo. 1959 9 sheets-sheet s 2 /NVENTE 5f w 'n Pie/:ard J'. LaMarma l/erne H- #Vf/son fforh e v United States Patent Oli ice 3,116,410 Patented Dec. 31, 1963 3,116,410 SIMPLE GENERAL PURPOSE DIGITAL COMPUTER Richard J. La Manna, Hanover Township, Verne H. Wilson and Vincent T. Pogorzelski, East Orange, and Mark Pivovonsky, Lyndhurst, NJ., assignors to Monroe Calculating Machine Company, Orange, N J.
Filed Nov. 30, 1959, Ser. No. 856,183 41 Claims. (Cl. 23S-157) This invention relates to general purpose serial cornputer apparatus. More particularly, in an important aspect, the invention relates to a digital computer having at least two synchronized information circulating loops.
In the present invention, the relatively short or minor loop is synchronized to the relatively long or major loop in an epicyclic timing relation. The computer operating cycle is based upon the recirculation of the minor loop and is divided into two phases. During the first of these one of many command codes is extracted from the minor loop. During the second of these phases the extracted command is executed upon a serially presented operand also circulating in the minor loop and, in certain instances, upon a second operand concurrently presented in serial fashion by the major loop.
This computer is characterized by a displacement type of relative addressing. In this type of addressing the address of the group of commands next to be executed by the minor loop may be a predetermined distance removed frorn the location in the major loop of the group of commands last transferred to the minor loop. Alternatively this address may be determined by specifying the amount of the displacement of this next group along the circulating major information loop relative to the current position of the minor loop.
Using relative addresses achieves a high eliciency in the use of equipment by avoiding the waste of providing the equipment necessary for searching throughout a serial memory for absolute addresses. Thus, in accordance with an important feature of our invention, addresses are absent in certain operations involving transfers from the major to the minor loop.
THEORETICAL DESCRIPTION Our invention discloses a computer which is intermediate on the one hand between a Turing type of computer, which by operating in a bidirectional serial manner on one binary bit at a time, uses a minimum of equipment at the expense of very slow operation, and on the other hand, a large scale stored-program parallel type of computer, which achieves very high speed operation at the expense of voluminous and complex equipment. Because of its efciency, our computer is capable of performing calculations involving business and scientic general purpose programs economically at speeds compatible with electromechanical input and output devices. At the same time computers in accordance with our invention require far less equipment than other general purpose computers. In addition, our invention teaches a computer organization which permits the stored program corresponding to a given computational purpose or format to be injected into the computer by a separate program loading device which can be shared by many such computers with consequent economy.
A basic principle of our invention is to use a minor circulating memory loop of reference information in such a Way that the minor loop at any instant contains both a group of operating instructions and multiple digit accumulator information. Our invention, in one preferred embodiment, then divides the time of circulation of this minor loop into two phases or half cycles during the first of which, designated precess time, one of the instructi'o`ns of the group is shifted out of this minor loop into an instruction register. During the second phase, designated execute time, the remembered instruction may be executed. Due to this automatic precession of the in structions, each instruction of a relatively large group may be executed in sequence without need for further transfer of instructions from the major loop until the entire group of instructions in the short, or minor loop is exhausted.
In this way of philosophy of opportunity is applied. Thus, while the main memory loop is moving toward the location of an eventually needed piece of stored information, the short loop is processing other pieces of stored information which lie on the way to the indicated location. The results of this last noted processing are deposited at relative positions known to the programmer. This may be considered to be an addressless type of computing which ows from the continuously active two-phasc minor loop in accordance with our invention.
A computer in accordance with our invention may be designed to execute directly as micro-operations such operations as storage to accumulator and accumulator to storage transfers, subtract, add, various jumps and conditional operations, and the input and output of a single character. A micro-program instruction corresponds to the performance of a single such operation upon every bit of a following word. It a micro-program instruction is modified during an execute cycle, the modified instruction advantageously may be executed during the remainder of the cycle.
Other macroscopic operations, such as multiplications, input and output of entire words, binary to decimal and decimal to binary conversions, and other specic operations that may be required for a particular application are programmed in terms of the foregoing primary microprogram operations.
Toward further minimizing equipment complexity, structures in accordance with our invention, employ time sharing of components made possible by the division of the circulation of the short circulating loop into the separate time phases of precessing and executing.
COMPONENT PARTS The elements of our invention include the two foregoing memory loops each with a reading station and a writing station. In the long or general storage loop, the writing station is downstream from the reading station so that information at a given location may be processed and replaced with the result of the processing at the same location. Conversely, in the short loop the reading station is advantageously downstream from the writing station. Thus, the information in the short loop advantageously may be continuously regenerated. These loops are preferably in the physical form of tracks on the surface of a magnetic drum, but, within the scope of our invention, they may also comprise a magnetic tape, a shift register, or any type of self-regenerating delay line.
These loops circulate information consisting of words each composed of a fixed number of binary bits. These equal length words may be either an instruction word, or a number word.
ORGANIZATION OF INFORMATION In one preferred arrangement in accordance with our invention, information is stored in a channel having a xed number of bits. The information stored in such a channel, by way of example, is divided in'to a xed number of storage sectors, say seventy-nine such sectors. Advantageously this number is an odd number as later appears. The information of such an illustrative sector is further divided into groups. Illustratively, each sector contains a single binary indicating bit, a flag bit, and a word having thirty-two additional binary bits. Each word is further subdivided into a plurality of subgroups which together include the remaining bits assigned to a sector. These subgroups may, for example, have informational significance such as the designation of arithmetic operations to be performed or, simply, the value of a number.
TIMING Clock means are provided to signal bit time and end of word time, and these times are lnade to be simultaneous for both circulating loops. End of word time is chosen to be coincident with the last bit in each word, and is conveniently designated taz for the above example, where tu is designated ag bit time.
A feature of our invention, as noted above, is the provision of an odd number of words or sectors in the general storage memory loop, and an even number of words (preferably two) in the short instructions-accumulator" loop.
The consequence of this provision is that on successive circulations of the general storage loop, a given general storage sector will correspond in time alternately to the rst, or instruction word, and then to the second, or operand number word, being circulated in the two-word instructions-accurnulator loop. It is this feature which allows any word in the general storage loop to be read into the short working loop either as an operator (instruction group), or as an operand (number). Thus, when it is desired to obtain a new group of instructions from the general storage loop, a revolution of the general storage loop is selected in which the desired new instruction word will correspond in time to the instruction part of the short loop, whereas when it is desired to nd a new operand (number word) from general storage, a revolution of the general storage loop is selected in which the desired number corresponds in time to the accumulator word in the short loop.
OPERATION OF THE COMPUTER In accordance with our invention, during normal precess time, a word is read from the accumulator-instructions loop and is permitted by suitable gates to flow through a shift register having as many stages as there are bits in an instruction. The output of this shift register is then written back into the short, two word loop.
At the end of word time of this precess phase U32), the gate leading to the shift register is closed and the shifting operation is stopped thereby stranding the last (highest order) command in the shift register. Due to this diversion from simple regeneration (by passing through the shift register), and since each shift occupies one bit time, the original instruction word under discussion will be delayed at the end of word time relative to the short loop by the length of one instruction. lt may be seen that each time this process occurs to an instruction word its information content is circulated or processed within the word by one command in the direction of higher order.
A single stage binary counter (ip-flop) is provided to count the end of word time signals and thereby to provide the two logical signal levels dividing the computer time into precess time and execute time. When this counter signals that precess time is completed, the proper gates are operated to decode the instruction left in the foregoing shift register and to execute this instruction on each of the bits of the following number word read from the instructions accumulator loop, or from the general storage loop, or both, depending on the command.
For those commands requiring arithmetic processing, a simple binary adder-subtractor is provided having two inputs, one output, and a carry-borrow memory flip-flop which may be interrogated or reset by the logic of the system. When this memory flip-flop is not being used for arithmetic purposes, it serves the important function of being an independent, conditional reference llip-op for decisions; this is in contradistinction to the usual computer practice of using an accumulatoroverflow as the condition for a decision` Simultaneously with the execution of the order, the results of logical operation being performed may be recorded, bit-by-bit, either back on the instructions-accumulator loop, or on the general storage loop.
It is to be noted that, in accordance with a feature of our invention, all recording (writing) operations in our invention are of the type which automatically erase whatever previously recorded bits may have existed on the loop medium. Thus, a previously stored word is replaced with a new word.
INSTRUCTION WORD STRUCTURE In a preferred embodiment of our invention, instruction signal combinations, i.e. information subgroups for designation of operations to be performed, are of four bits length. These four binary bits give rise to sixteen available instructions which may, illustratively, be allocated between twelve arithmetic operations and four displacement addressing operations. These displacement operations are those in which, for special programming purposes, the normal application sequence of operating instructions to stored information is interrupted and a new set of operating instructions introduced.
These instructions are thus executed in sequence under control of the word counter which repetitively alternates between states representing precess time and execute time, thereby bringing in a new four bit instruction from the current instruction word on each precession.
One of the above noted sixteen instructions, a skip instruction, is represented by a combination of all zeroes. This skip instruction does nothing except to regenerate the number in the accumulator portion of the short loop. The importance of this instruction appears hereafter.
After each instruction is executed, the instruction is cleared from the shift regitser (replaced by binary zeroes) so that the process of order code precession is characterized by a discontinuous march of the low order Hag bit followed by these zeroes automatically injected from the shift register towards the high order end (last in time) of the instruction word.
A simplifying feature of our invention is the choice of a binary one for the flag bit in combination with the rst three automatically following zero bits to comprise the four bit instruction for an automatic fixed length displacement command, designated jump. After the last order written by the programmer has been executed, a 1G00 signal combination articially generated in the ninth precession enters the shift register and is decoded to effect a xed displacement addressing operation. A new instruction word is transferred from the general storage without wasting program space in the original instruction word for a jump order. Obviously, the programmer must be sure to provide a new instruction word in general storage at the fixed jump distance from the end of the last instruction written.
It will be seen that this device of so choosing the pattern of a command code that its topology produces useful results is one illustration of a powerful method taught by our invention, and, as will be subsequently disclosed, used to convert one command code into another on the basis of a decision.
ADDRESSING OPERATIONS The displacement addressing of our invention is accomplished by what we designate a jump delay instruction. When this instruction is included in an instruction word, room must be left to include the binary displacement address at the low order end of the instruction. This binary address number thus replaces one or more tetrads of bits which would otherwise be instructions. On decoding the jump delay, the logical network registers this fact n a mode" flip-flip memory, and then does nothing for one execute period except to regenerate the information in the short loop. On the following precess period, thc normal instruction word precession (through the shift register) is changed and instead,
the remainder of the instruction word, which now comprises the binary number representing the displacement address, passes through one of the two inputs of the subtracting unit. The subtractors output is stored in the short instructions-accumulator loop. In this way, by properly setting the other input of the subtractor, the address number suffers a diminution by one unit each time the short loop undergoes one regenerative circulation, namely, each two word times. Meanwhile, the number word in the accumulator is being regenerated intact during the entire jump.
When the address number in the instruction word has been reduced to zero, this fact is signalled by the condition of the borrow flip-flip, and this combination of conditions causes the logical network to terminate the jump with the reading of a new instruction group word from the general storage loop into the instruction word interval of the short loop.
ln this manner, a relative address has been counted down" to permit the computer to select from general storage a new set of commands as directed by the programmer. It will be seen in the following specification that during this count down, the flag bit has the additional and important function of marking the beginning of the address number regardless of the composition of the instruction word.
BRANCHING OPERATIONS Branching in our computer may be accomplished by a jump conditional order. In this case, the xed-displacement (jump order) seeking a new series of instructions is made to be conditional on the state of a condition indicator flip-flop, which can be made to be a memory for the success or failure or overflow of a previous operation. If this condition indicator memory fliptlop is not in the predetermined state to cause a jump, the logic is arranged to change the jump conditional command into a skip which regenerates the accumulator word in the short loop and proceeds to the next instruction in the current series.
It may be seen that a jump conditional order followed by a jump delay order, in which the delay or relative address is made to be equal to two circulations of the general storage loop, can be used to set up an iterative test or interrogation routine. The routine will continue until some external condition changes the state of the conditional reference flip-flop, thereby causing a jump and the starting of a new sequence of instructions. This interrogation routine is useful for entering input information if the computer must wait until it receives a keyboard signal, or in providing output information if the computer must also wait for a feedback signal from the printer signifying that it is no longer busy and is ready to receive a new print command.
SPECIAL CONSIDERATIONS It is n characteristic of our invention that the epicyclic circulation is continuous, and, therefore, there can be no stop instruction in its commands. This fact makes the starting procedure of the computer an essential part of our invention, rand it will ibe seen that we disclose a novel method of starting up and shutting down the machine in such a way as to enter and leave the epicyclic circulation state without destroying the stored program. The ag bits in the general storage track play an important part in this operation and in the loading operation, by providing an index signal to locate the initial instruction` word.
Another feature of our invention is the leap cornmand which inhibits the word counter action at 132, thereby effectively interchanging preceiss time and execute time, and making the number word in the short loop an instruction word and vice versa. This order is useful in abridging the two circulation epicyclic correspondence period, and in using instruction words which have been built up as a number.
The remaining elements of our invention comprise an input register, an o-utput register, a tetrad counter to distribute input `and output signals between parallel and serial presentation, a mode flip-Hop to distinguish between arithmetic and displacement type commands, and at least one delay flip-flop for giving early or anticipated access to the short circulating loops.
The gating and reset logic of the system is thus arranged to control the information flow between the reading and writing heads of the two loops, the shift register, the add-subtract net, and the input and output registers in accordance with signals received from the clock means, word counter, shift register, conditional reference ipflop, tetrad counter, and mode ip-llop. In addition, the input and output devices are provided with switches to signal to the logic their availability.
Thus, in accordance with the invention, structures are provided for achieving related principal objects. These objects are to eliminate quiescent time in the utilization of computer elements thus to simplify, to render more economical, and to heighten ithe reliability of computer structures.
In achieving these objects it is a feature of the present invention that there are provided structures having a relatively short circulating information loop which contains an operand which is subjected to continuing modification in accordance with a group of instructions sequentially circulating with `the operand in the information loop.
It is an additional feature of structures in accordance with the invention that instructions are precessed out of a working information loop into a staticizing register and then executed upon an operand following the same loop.
lt is a further feature of the invention that there are provided structures having a series of micro-program instructions sequentially circulating in a loop with an accumulator number so that each micro-program instructions, in turn, is executed upon the accumulator number.
A further feature lies in the provision of structure for the relative addressing of gro-ups of instructions in `accordance with both programmed and unprogrammed instruction signal combinations.
Another feature of the invention lies in a circulating information loop in `which an operand may be placed for being subjected to continuing modification in accordance with a group of instructions sequentially circulating with the operand in the information loop.
An additional feature is the provision of structures in which instructions `are alternately processed out of a working information loop into a register and then executed on an operand following in the same loop.
A still further feature of the invention is the provision of structures for accomplishing fixed displacement addressing in response to predetermined instruction signal combinations which selectively initiate transfers of information from a relatively long loop after a predetermined number of cycles of a relatively short loop.
An additional feature is the provision of logical branching paths upon the displacement of a relative address which is partially conditioned on the state of a memory device.
Another feature of the invention is a momo-ry arrangement in which a specific word location in a major loop corresponds in time to la given word location in a mino-r loop only on `alternate circulations of the major loop.
These and other objects, features, and advantages of the invention, both as to its organization and method of operation, will be better understood by consideration of the appended claims and from `the following description considered in connection with the accompanying drawings in which several embodiments of the invention are illustrated yby way of example. It is expressly to be un` derstood, however, that the drawings are for the purpose of illustration and description only, and are not intended as a denition of the limits of the invention.
FIGURE l is a generalized illustration, partially in block tand partially in isometric view of a preferred oomputer according to the present invention;
FIGURE 2 is a representation of a portion of the memory channels of the computer of FIGURE 1, showing the time correlation between cells of the timing channel and the corresponding cells of the A-I register;
FIGURE 3, consisting of FIGURES 3a and 3l, is a series of idealized information ow diagrams, each corresponding to a different micro-program operation;
FIGURE 4, consisting of FIGURES 4a and 4b, is a timing diagram illustrating the operation of an input and output counter;
FIGURE 5, consisting of FIGURES 5a through 5d, is a more detailed diagram of the reading and writing circuits and the relay circuits of the computer of FIGURE l;
FIGURE 6, consisting of FIGURES 6a through 6p, is a series of block diagrams showing in detail the various gating networks included within the logical network of FIGURE 1;
FIGURE 7, consisting of FIGURES 7a and 7b, is a pair of flow diagrams of a calculator program written for the computer of FIGURE l; and
FIGURE 8 is a block diagram illustrating information ow in an illustrative structure in accordance with the invention.
Turning now to the gures, FIGURE 1 is a diagram of a computer according to the present invention. In a pre ferred embodiment of the present invention, a magnetic drum 20 driven by a motor 2.1 has a magnetizable surface which is divided into four drum channels. A timing or clock channel 22 is recorded, containing 79 evenly spaced binary word sectors, each having 33 cells. Each cell of u channel is capable of storing a single bit of information represented by a predetermined state of magnetization within the cell. Each of the 79 sectors of the timing channel 22 corresponds to storages for 33 binary bits.
Within each sector of the timing channel, 32 cells contain binary representing bits to mark the rst 32 bits of each word and the last cell of each word stores a binary 1" representing spot to identify uniquely the last bit of each word, which is designated x32. The individual magnetic spots within the timing channel 22 are detected by a T, timing read head 24 `which generates electrical signals and applies them to a timing pulse generating unit 26, which is described in greater detail in connection with FIGURE a below.
A second track on the drum 20 is used to create a two word recirculating memory loop which is designated the Accumulatonlnstruction channel 28. The Accumulator- Instruction or A-I channel 28 is used to store a rst binary word representing a number, and a second binary word representing a series of instructions corresponding to operations to be performed upon the number. As will be pointed out in greater detail below, the circulating two word memory loop or register, mechanized by the A-I channel 28, uses only 64 storage cells on the drum corresponding to 64 cells of the timing channel 22. The additional two storage cells necessary to store two 33-bit groups are provided by two static storage elements, more fully described below. Associated with the A-I channel 28 is a writing transducer, the A-lw write head 39 and, separated from it in the direction of drum rotation by a distance equivalent to 64 cells is a read transducer, the A-Ir read head 32.
The remaining two magnetic channels of the drum are used as information storage channels and are designated general storage 1 or GS1 34 and general storage 2 or G82 36. As noted heretofore, general storage channels each contain 79 binary sectors each of which contains a single ag bit though in other embodiments of the invention a different number of Hag bits may be advantageous. Within each sector, bits other than iiag bits make up a word which is further divided into subgroups. The subgroups may have an instruction significance or a number significance. In conventional partance they are instruction words or number words. For clarity of discussion, however, these are considered as "subgroups of the larger "vt/ord. Each of the general storage channels 34, 36 has an associated read transducer, a G51r read head 38 and a G52r read head 4i), respectively. Information is entered into the general storage channels by means of a pair of write transducers, a GSIw write head 42 and a G82W write head 44, respectively.
The writing heads of the general storage channels are displaced from the reading heads in the direction of drum rotation by a distance equivalent to six word sectors plus one ccll or 199 cells of the timing channel 22. This arrangement permits information from the general storage channels to be read from a given sector, operated upon in the A-I register and returned into general storage, in the same sector replacing the former information by processed information. It is therefore a relatively simple matter to update" stored quantities. The extra cell between the read and write heads is necessitated by the fact that the general storage channel operating circuits include a reading staticizer which introduces a one-bit delay that must be accounted for.
As will be explained in more detail below, conveniently the words in general storage have, as a low order or least significant bit, a binary l representing signal which is designated a at bit. More specically, in this embodiment, cach sector of the general storage, except for one, has this binary "1 ag bit. However, in a predetermined word of one of the general storage channels, the least significant or flag bit position is occupied by a binary 0. It is this 0 flag bit that serves as an index pulse to synchronize the circulating loops of the computer, and control the START and FILL operations.
A logical gating network 50 is shown only in block form in FIGURE l but is explained in greater detail in connection with FIGURE 6, consisting of FIGURES 6a through 6p, described below. The logical network 50 includes the read and write amplifiers and staticizers necessary for operation of the drum and its associated read and write transducers. Also included within the block, are all of the and and or gates, amplifiers, and inverters, which are necessary to execute the various logical functions during the operation of the computer'. These gate circuits receive signals from the drum transducers during certain time intervals, operate upon them, and provide signals to the drum transducers representing the result of a predetermined operation upon concurrently presented information signals. The predetermined operations correspond to the micro-program instructions presented by the A-I channels in a prior word time.
Shown, arrayed about the logical gating network 50, is a plurality of flip-flops, indicated by a plurality of individual rectangles. Each rectangle contains an alphanumeric identifying designation as well as a reference character. Each flip-fiop may be a conventional bistable multi-vibrator, well known in the art, having two inputs and two outputs corresponding respectively to each of the stable states of the Hip-Hop. Each ilipflop receives control signals upon conductors from the logical gating network 50 for placing the flip-flop in either a set or l representing state or a zero or 0" representing state, and which are designated respectively, by the alpha-numeric designation of the flip-flop preceded by the letter S for set and Z for zero. The set output conductor has the alpha-numeric designation of the flip-flop and the zero output conductor has the alpha-numeric designation primed.
An L2 flip-flop 52 is connected to the logic network and is normally used as a read staticizer for information coming from the A-Ir read head 32. An F2 ip-tlop 54 and an F3 flip-flop 56 are connected to the logical gating network 50 to be used, at different times, as staticizers for the information read from the general storage read heads and, during certain operations further described below, the F3 flip-flop 56 is connected in shift register fashion to receive the output of the L2 Hip-Hop 52. Four ip-ilops, F4 58, F5 60, F6 62, F7 64, are, during certain operations, connected together during the precess period as a four-bit shifting register and, as such, staticize four-bit instructions to control operation upon the succeeding accumulator number.
An F1 Hip-Hop 66 is used primarily as a carry-borrow Hip-flop in arithmetic operations. In other operations of a conditional nature, and in input and output operations, as will be described below, the F1 ilip-op 66 serves as a supplementary memory'. Another Hip-hop, the FS flipilop 68, may be considered a two-counter which assumes alternate stable states in successive word intervals. The F8 flip-flop 68 can therefore be used as a basic reference, the state of which marks the precess and execute periods, and therefore identifies the signals appearing at the A-Ir read head 32 of the A-I register as number signals or instruction signals. In one state, F8, the signals appearing are bits of an instruction word while in the other F8 state, the signals representing bits of number words are presented.
An F9 ilip-op 70 is another two-counter that remembers which of the two general storage tracks is in communication with the computer. Each time a track change is instructed, the state of the F9 ip-fiop is changed. The output of the F9 flip-Hop therefore can be used as a logical input for head switching as between the general storage tracks, F9" selecting GS1 and F9 selecting GSZ.
A [p32 flip-op 72 is provided to identify uniquely the time interval corresponding to the last bit time of each word. The tp32 flip-flop 72 is in a rst stable state, w32', during the first 32 bits of each word, but during the last or 33rd bit, designated 132, the hip-flop assumes its other stable state, tp32. The tpgz flip-flop 72 can also be used to identify the lowest order or ag bit at time tu lof the next succeeding word. Inasmuch as the recirculation path of the A-I register normally includes both the L2 flip-op 52 and the F3 ip-fiop 56 to total 66 bits of storage, when the bit corresponding to tu is staticized in the F3 flip-flop 56, and the :p32 llip-ilop 72 is in its tpm representing stable state, the L2 flip-flop 52 will be staticizing the t or flag bit of the next word.
An input-output unit 80 is provided to enable the computer to communicate with the external world. A plurality of input keys (not shown) are provided to generate a combination of binary signals on tive input conductors and a sixth control signal on a separate conductor. The logical gating network 50 applies a seventh signal to the input devices on a seventh control signal conductor. The gating network 50 provides a combination of signals to the output device on iive output conductors and a sixth control signal on a separate conductor. The output device returns a control signal on a seventh conductor. A multi-conductor input cable 82 transmits the signals between an input portion 84 of the input-output unit 80 and the gating network S0. A multi-conductor output cable 86 transmits signals between the gating network 50 and an output portion l88 of the input-output unit 80.
A pair of relays R1 90 and R2 92 are provided to facilitate the starting and stopping of the computer. The manner of interconnecting the relays is shown in greater detail in connection with FIGURE d below. Logic provided within the logical gating network 50 is responsive to signals from the relays to control the energization and de-energization of the electromechanical and electronic components associated with the computer.
There is shown in FIGURE 2, an instruction subgroup of digits 102, at the read point of the AI register and a number subgroup 104 in one of the general storage channels aligned with a portion 106 of the timing channel 22 corresponding to one sector. In the preferred embodiment of the invention, as was pointed out above, each sector is 33 bits in length having a flag bit in the lowest order or to position and, in the case of number 10 words, 32 binary digits. An instruction word contains a flag bit and up to eight, four-bit instructions. In the case of certain relative address operations, however, an instruction word has less than eight instructions, with the remaining group of digits representing a binary number corresponding to a relative address.
As shown, the cells of the individual channels are arranged in vertical alignment representing concurrence in time only. As is well known to those familiar with the art, the individual cells in a given channel may have any physical position with respect to the cell-s of a different channel so long as the read and write transducers are positioned to present signals synchronized in time. Furthermore, the read transducers are indicated as windows on their respective channels and, in the portion of the A-I channel, the two read staticizers, the L2 and F3 flip-flops are also shown as windows on the channel.
It is therefore apparent from FIGURE 2, that the ilag bit signal of the general storage channel words are available at the general storage read transducers simultaneously with the availability of A-I channel word flag bit signals at the L2 flip-Hop. This correspondence in time permits the output of the F2 ilip-op, when used as the general storage read staticizer, to be combined with the output of the F3 flip-flop being used as an A-I read staticizer stage.
As was discussed earlier, the operation of the computer is divided into two basic periods corresponding to the half cycles of the A-I circulating register. `Each half cycle is one word time in length, alternate half cycles being designated precess and execute word times. During each precess word time, four bits of an instruction word, representing a single instruction which corresponds to an operation to be subsequently performed, are extracted from the instruction word and are staticized in a four-bit register. During the succeeding execute Word time, the staticized bits control the disposition of the number presented by the A-I register. In the execute half cycle, the A-I register may be considered as an accumulator register, storing a number to be operated upon. In order to set up each instruction in the four-bit staticizer, four zeroes are inserted in the A-I register to precess the remaining instructions toward the most significant end of the instruction word. After eight complete cycles of the A-I register, an instruction word consists only of the ag bit in the most significant bit position followed by 32 zeroes.
With instruction signal combinations of four bits each, 16 different program instructions are available to the computer. The selected micro-program instructions of the preferred embodiment are set forth in Table 1 below. Table l has been arranged to list each instruction by name, an abbreviation, a binary code number and the declmal number equivalent of the binary code.
Table 1 Name Abbre- Binary Decimal viation Codo Number Skip SK 0000 0 Transfer to General Storage TS 0001 1 Transfer to Accumulator TA 0010 2 Extract or Leap: F1=Extract, Fl'= Leap E/L 0011 3 OUT 0100 4 p 1N 0101 5 Subtract` (.rcnernl Sto fc Numberfroin Accumulator Number SUB 0110 G Add General Storage Number t0 Aecum ualtor Number ADD 0111 7 Jump 1000 8 Change Track Delay 1001 9 Jump Conditional 1010 10 1011 11 1100 12 1101 13 lll() 14 Leit Shift. (Multiply by 2 llll 15 As seen in FIGURE 2, an instruction word is arranged with the `most significant bit of the first instruction located in the last or t32 bit position. It is understood that the flow of information is such that the recorded information moves to the right, passing under a stationary read transducer A-Ir indicated by the properly designated Window at the right side of the drawing. In the instruction word 102, the first instruction to be executed includes four bits, in, i12, im and 14, stored respectively in the t32, x31, t3() and f2s bit positions. Similarly, the bits of the second instruction, i2, through i2., are stored in cells corresponding to t28 through 125. The bits of the last instruction, 1'81 through i8', are located in the cells t., through t1.
If an instruction word includes a delay type instruction which re-quires a jump to some relatively remote location in general storage, the instruction requiring this transfer is followed by a binary number which represents a relative address of the word being sought. This binary number is located at the least significant end of the instruction word and occupies as many cells as are necessary to specify the relative address.
A typical number word 104 contains a flag bit at the low order end `and a plurality of binary digits which may form the number. The most significant digit position, t32, of the number word may be interpreted as a sign bit at the programmers discretion.
Eleven of the sixteen primary operations, set out in Table 1 above, may be considered arithmetic instructions in that they utilize the accumulator. These eleven consist o-f the following: transfer to general storage (TS), transfer to accumulator (TA), extract-leap (E/L), output (OUT), input (IN), subtract (SUB), add (ADD), clear (CL), right shift (RS), add conditional (ADDc), and left shift (LS). Note that the transfer to general stor- :age command is the only instruction which records new information in general storage. Also, live of these instructions, namely, transfer to accumulator, extract, subtract, add, and add conditional, utilize the general storage read head as the source of the second operand.
Four of the sixteen primary commands serve as displacement addressing instructions. These are the jump group, namely, jump (I), change track (CT), jump conditional (Jc), and jump delay (Irl). These instructions operate on the instruction portion of the A-I channel during precess time in contradistinction to the foregoing arithmetic commands which operate on the accumulator portion of the A-I channel during execute time.
The remaining one of the sixteen primary operations is the skip instruction (SK), so titled because it leaves everything unchanged.
The most effective approach to the description of our invention lies in the detailed examination of the role played by the nine basic system flip-flops (F1 to F9) and the two retiming flip-flops, 1.2 and 11232, in effecting the sixteen primary operations.
In the following discussion, FIGURES 3a-3k are especially useful in correlating the time sharing functions played by these flip-flops during each instruction, and in following the information flow paths for these operations.
The F1 Hip-flop 66 is the general purpose conditional memory which is used during the relevant arithmetic operations as the carry-borrow flip-flop for the addersubtractor unit. 1n general, F1 remains static during precess word times (except during the jump delay and change track commands) so that in effect it carries information forth from a previous execute word time to a subsequent one.
Thus all the conditional operations (add conditional, jump conditional, extract-leap, jump delay, and change track) are conditioned on the information content of F1, remembered from the previous operation. It should be noted that the conditional part of the add conditional, extract-leap, and jump conditional commands is whether or not they are to be executed, whereas in the jump delay and the change track commands, the condition of the Fl flip-flop is not a condition to the execution, but instead modulates how much displacement will take place as determined by the delay address numbers in the given instruction word. For programming convenience, F1 is usually referred to as the carry flip-flop (designated K flip-flop) although it actually serves as a multi-function conditional register.
During execute word times, tiip-tlop F2 54, serves as the general storage read flip-flop in those operations where a general storage operand is needed. During the input and output instructions, F2 teams up with the F3 liip-liop S6 to serve as a tetrad counter to identify time intervals to to t3. In other operations, F2 serves as a counter Hipflop to designate I@ intervals during execute word times. In the precess word times, however, F2 plays an entirely different role. Here it serves as a mode flip-flop to designate whether the computer is undergoing a normal precess word time or if one of the four jump commands (jump, jump conditional, jump delay, or change track) is being performed. Note that the F3 rather than the F2 ip-op serves as the general storage read tiip-iiop when a new instruction word is being transferred from general storage to the instruction portion of the A-I channel during a jump type operation. Thus it is seen that F2 represents the general storage read flip-flop for number words (during execute word times), whereas for instruction words (during precess word times) F3 represents the general storage flip-flop, when necessary.
During normal precess word times, F3 represents a second A-l channel read flip-flop, in series with the L2 flip-flop. F3 serves this same role during execute word times for most of the commands. During input and output, F3 in conjunction with F2 serves as a tetrad counter (from to to r3) as previously mentioned, but in general F3 represents an A-l channel read flip-flop whenever it is necessary to use the accumulator as an operand, regenerate the accumulator, or normally precess the instructions.
The F4 to F7 flip-flops, 58, 60, 62, 64, represent the four flip-flops in the command shift register, under control of the F8 tiip-tiop 68. F8 may be considered as the Word time counter, in that its sole function in the computer is to count sector pulses at t32. As such, it serves as a fundamental reference flip-flop.
During F8 (precess word time) the command shift register advances the instruction Word from the A-I channel read head into the A-I write head, as is best noted in FIGURE 3b. The high order tetrad of the instruction word remains locked in the command shift register as F8 ends and F8' (execute word time) begins. During F8' this tetrad is decoded from the command shift register as one of the sixteen primary mirco-program instructions.
During the precession in F8, all the tetrads in the A-I channel are precessed up one tetrad towards the high order end. Thus the second highest order tetrad before the precession now becomes the new high order tetrad in the instruction word portion of the A-I channel. As such, it is in position to be precessed into the command shift register during the next F8 precession, where it will be used as the next operation code. It follows that the command shift register normally never advances during F8 since the operation code is being decoded from it. Conversely, the command shift register always advances during a normal F8 precession (a precession other than one involving a jump type instruction) in order to set up the next primary operation code.
The F9 flip-flop 70 represents the change track flipfiop, and so controls which of the two general storage tracks is selected for reading or writing. This control function is best seen in connection with FIGURES 5a and 5b, described below. The only time that F9 changes is during a change track instruction (CT). The change track instruction is identical in all respects to the jump delay instruction, except that the F9 {lip-Hop is triggered to its complementary or opposite state as the instructions name implies. Thus the change track command is really a member of the jump type (displacement address) instructions identical in all ways to the jump delay except for its one special duty. F9 serves one additional purpose of serving as a fth channel of output to the typewriter or printer thereby giving the programmer a choice of 32 output characters rather than the 16 provided by the four binary digits.
The logical gating diagrams, FIGURES 6er-6p, comprise a detailed schematic of the logical gating network 50 from which those skilled in the art can reproduce an operating model of our invention.
In the following descriptions of the logical operations of our computer for each of the sixteen instructions, reference is made to Table l which sets forth each of them. Particular note `should be made in each case of the disposition of the:
(l) The K (F1) flip-Hop (2) General storage word (3) The number word in the accumulator (l1-132) (4) The low order or flag bit of the accumulator, to
SKIP (SK) (0000) K, general storage, and the number in the accumulator remain unchanged. The low order bit of the accumulator is reset to zero.
Translating the foregoing definition into computer logic terminology, it is observed that all that is necessary to achieve this command is to regenerate the number in the accumulator. F1 (the K flip-Hop) and general storage will automatically remain unchanged unless we specifically write some logic (i.e. provide gates) to change them. The How path for this operation is set out in FIGURES 3a' and 5c, and the specific gate path is shown in FIGURES 6er-6p.
The regeneration of the accumulator requires the information word to pass from the A-Ir head, 32, back to the A-IW head 30. Filling in the `gates for this path from FIGURES 6a-6p, signals pass from the A-Ir head 312- LS-SLZ [FIGURE 611], G8-G12-G17-F30-SF3 [FIG- URE 6bl, G53-G63-L1 [FIGURE 6n], G20-L7 [FIG- URE 6d], to A-Iw head 30.
To illustrate a detailed tracing of such a logical gating path, it may be seen that, starting in FIGURE 5b the A-Ir head 32 leads through a non-inverting amplifier 110 to generate a signal on conductor L8 and through the inverting amplifier 112 to generate a signal on conductor L8'. Thus if a l is read by A-Ir head 32, a relatively high or l representing signal will appear at clock time C1 on lead L8 While if a 0 is read by head 32 a relatively high or 1" representing signal will appear within bit time C1 on lead LS'. Turning now to FIGURE 6h, these leads are shown applied to control the setting of the L2 flipliop, 52. It is to be understood, at this point, that all setting and zeroing inputs to the Hip-flops are clocked by timing pulses C1 which are not necessarily shown in the figures. The state of output conductors L2 and L2' is then high or not depending on the presence of a l or a 0 respectively at the A-Ir head 32 `during the previous bit time. Gates G8 and G12 are found in FIGURE 6b, and it may be seen that the foregoing L2 signal, together with a signal from gate G8, and the F8' (execute time) signal are all necessary to pass through the and gate G12 to reach or gate G17. The signal path then oontinues through the network designated F30 to control the setting of the F3 tiip-op 56 one bit time later, thereby generating either F3 or F3'. Gate G55 in FIGURE 611 is dependent on gate G53 and on the foregoing F3 signal, together with A10, to continue through gate G63, inverting amplifier 114 to L1'. The signal L is applied in FIG- URE 6d to gate G20, the output of which generates a signal L7 by a rst inverter 116 and signal L7' by a second inverter 118. Returning to FIGURE 5b, the path leads to the A--lW head 30 via gate G1121, one shot pulse generator 120 and non-inverting amplifier 122 in the case of an L7 or 1" representing signal to be recorded or via gate G122, one shot 124, and non-inverter 126 in the case of an L7 or 0 representing signal.
The F5 diode in G8 and the F6' diode in G53 allow this path to operate. It should be noted that the GS- G12-Gl7 part of this path does not regenerate the low order flag bit (tu) of the accumulator. The reason for this is that the F30 net (FIGURE 6b) is timed one bit early because of its position in the A-I register circulation loop (this bit represents the `unit delay in the F3 ipflop). The F3 llip-op contains the most significant bit of the instruction tetra-d, which is a 0" at x32 of the previous F8. The low yorder bit of the accumulator number is in L2 :and the F30 net. At 32 of F8, no gate of FIG- URE 6b can transfer the contents of L2 to F3. The F3 flip-flop therefore copies its contents into the writing circuits, thereby resetting the IOW order bit position to 0, thus preventing the regeneration of the low order accumulator bit.
TRANSFER TO GENERAL STORAGE (TS) (0001) (1) The K flip-flop and the number in the accumulator remain unchanged. The flag or low order bit of the accumulator is reset to i). T'he number in the accumulator is transferred to general storage thereby replacing the number previously stored there as is shown in the ow path of FIGURE 3C.
Aside from the additional function of transferring the number in the accumulator to general storage, this command is identical to the skip command. Thus the previous description of the skip command serves equally well with respect to K, the regeneration of the accumulator, and the `loss of the low order accumulator bit. The accumulator regeneration path given for `the skip command is identical for the transfer to general storage command. The path from the A-Ir read head 32 to ZF3-G18-L10'- (G114 or C117) and A-Ir read head 32 to SF3-Gi19- L11-(G114 or G117) represent the flow of information signals being transferred from the accumulator to the general storage write circuits. The rst of the two paths represents that taken by the Os in the information word, `while the second represents the path of the ls.
In the transfer to storage command, it is to be noted that this is the only one of the primary commands which records into (changes) general storage, and that the general storage write heads, GS1w and GSZW, `42, 44 are six word times (plus 1 bit) behind the GS1r and GS2r read heads (38, 40) so that the `word replaced by this command is six sectors removed from the word concurrently presented by the general storage read heads, G51r and G52, 38, 40.
In decoding the logic for G18 and G19 in the path from the accumulator to the general storage write circuits, we nd the equivalent of an F2 diode ron each of these gates. The F2 flip-flop 54 is set to represent time intervals tlm or to' in the transfer to general storage command. Thus we see that the purpose of these F2 diodes is to protect the flag bit in general storage by preventing the transfer of the low order (to) accumulator bit to general storage.
TRANSFER T() ACUUULATGR (TA) (0010) (2) The K flip-hop and general storage remain unchanged. The low order bit of the accumulator is lost (reset to 0l). The number in general storage is transferred to the accumulator, and replaces the number previously there (FIG- URE 3d).
The source of the number in this transfer is a general storage read head (38, 40), which, as has been previously noted, is six sectors removed from the general storage write heads. The actual path of transfer from general storage to the accumulator is L12-(G1, C109, G2, G3)-G4G7-F20-SF2 [FIGURE 6u), G60-G63- L1' [FIGURE 611]. G20-L7 [FIGURE 6d). The F5 diodes in G3 and implicit in G1 through the A5 diode [FIGURE 6p] and the L12 diode in G109 and G2 [FIG- 15 URE 6a] allow the path to operate through G4 [F1G URE 6a].
As usual K and the general storage remain unchanged since no logic is written to change them. The F2 ipflop 54, which represents the general storage read llipflop in this command as is seen in FIGURE 3d, is reset originally as we enter the command. This means that is transferred into the low order bit position of the acetirnulator at to, and consequently the low order bit previously there is lost (reset to 0).
lItlXTRACT-LEAI (lil/L) (0011) (3) If the K Hip-flop is 1 as this command is entered, the extract order will be executed. lf K is 0" as this Command is entered the leap order will be executed.
In the extract order, K and general storage remain unchanged. The low order bit of the accumulator is lost (ire-set to 0). The number in general storage is logically multiplied by the number in the accumulator and the result is placed in the accumulator (FIGURE 3g).
K and the general storage remain unchanged for the usual reason that no action is taken otherwise. In com puter logic terminology, a logical multiplication is simply an and gate. In other words, it takes a ls bit in both of the operands to yield a ls bit in the logical product for any given position of the number. The and gate in question is gate GSS (FIGURE 6:1). The path of the general storage operand up to this gate is L12-(Gl, 6109, G2, G3)G4-G7F20-SF2 [FIGURE 6u), G53- GSS [FIGURE 6:11, while the path of the operand from the accumulator is LS-LZ [FIGURE 611], GS-GlZ-GIT- F30-SFI) [FIGURE 611], G55. The FS' diodes in G1, G3, and G8., and the L12 diode in G1109 and G2 allow these paths to operate. After the operands have merged in the and gate G55 to give the logical product, the path continues as G55-G63-L1 [FIGURE 6u), G20-L7 [FIGURE 6d] `As in the transfer to accumulator command, F2` once again represents the general storage read flip-flop and consequently is reset originally. Hence a 0 is transferred at to from F2 through` G53 to G55. With this input condition the output of G55 (representing the logical product at to) must be 0, and consequently, the low order bit of the accumulator is once again lost (reset to 0).
In the leap command, which is directed by the (00l l) code when 14:0, K and general storage again remain un changed and the low order bit of the accumulator is again reset to 0. However, the presence of this command at 132 of F8 prevents the immediately `following 132, end-ofword clock bit from performing its normal changing of F8 (precess time) into F8 (execute time). As a result, the accumulator word which normally would have been processed by some instruction beginning with FS suddenly is `regarded by the computer as a regular instruction word and `is precessed through the shift register to extract a new command. This reversal of roles between the current accumulator word and instruction word has the same eiect as if the drum had suddenly and discontinuously lost or gained one revolution. This, therefore, immediately gives the instruction portion of the A-I channel access to Words in the general storage track which it would otherwise have had to wait a full drum revolution to achieve. Conversely, the accumulator portion of the A-I channel is given immediately access to words on the general storage drum for which it would otherwise have had to wait the time of a full drum revolution.
In order `to perform the foregoing task for the leap order an or gate G? (FIGURE 6p) is provided as an input to and gate G75 which is enabled at t32. Gate G10? `is responsive to signals from the F3, F4, F5, and F6 flip-flops which store the command that will be in the command register at to. If this tetrad is F3', F4', F5, F6, and also if F1 `is set to "0 (i.e. Fl), then a leap command exists and circuitry should be provided to inhibit the triggering of the F8 flip-flop. This is accomplished by making the operation of the "and" gute G75 16 contingent on the presence of any one of. the de Morgan (primed) elements of the leap condition, Since the complement of an and expression is an or expression of the complemented individual terms, if either F1, F3, F-t, F5', or F6' is present at gate G107 at tgz, it means that the leap order is not present and the normal generation of F 8 is permitted. Conversely, if none of the above five signals is present, a leap order is present, and gate G is prevented from generating the auxiliary logical signal A14 (FIGURE 16p) which is applied `to the ZFS termi nal of the F8 flipllop 68, as seen in FIGURE 6g.
OUTPUT (OUT) (0100) (4) If the busy signal (E10) is present, K is reset to 0 and the accumulator remains unchanged. If the busy signal is not present K is set to l, and the low order tetrad (four bits from t, to t4) of the accumulator is printed and cleared to Os. The remainder of the accumulator regenerates and the low order (to) bit of the accumulator is cleared (reset to 0) regardless of the busy signal.
The first case to be analyzed in the output command will be where the busy signal E10 is found to be present. As shown in the foregoing leap command, it is possible to examine an incoming code before it is fully shifted into the command register, F4, F5, F6, F7, namely at tgz of F8 when the command is arrayed in F3, F4, F5, F6.
In the case of the output command, we anticipate it at {32 of the previous F8, and block the passage of the ls bit in the output operation code (0100) from F4 to F5 if we find the busy signal present denoting an unsuccessful output. Thus if the output command is unsuccessful (E10 present), it is automatically changed into a skip command as F8 changes to F8. Conversely, if F8' is reached and the output command is still present, we know that it is a successful (E10 not present) output. The G26 "or gate (FIGURE 6j) is the gate for transferring this l's bit, and the path of the signal is E10'G26-G27-SF5. If E10 is present, obviously the path cannot operate and the F5 flip-flop 60 remains in the reset or 0 state.
The normal path of information flow at this point in the command shift register is F4-G27-SF5. It will be observed that the G26 "or gate represents the only point where the busy signal, E10, enters the logic. Thus it may be said that the only function of the busy signal, if present, is to convert the unsuccessful output command into a skip command. The denition and explanation of the skip command, as previously described, serves to complete the unsuccessful output.
If the busy signal is not found present and the output command arrives intact in the command register at F8', the output is initiated. FIGURE 4a represents the timing diagram for the output operation. In essence, the F2 and F3 flip- flops 54, 56 combine as a two stage tetrad counter as set out in Table 2, below.
The F2 F3 final condition extends across the rest of the word time, but at the end of t3, the output command code is reset to a skip command as shown by the state of the F5 ip-op 60 in FIGURE 4a. At the same time the K flip-flop F1 is set to a 1 in order to serve future notice that the output command was successful. Thus we observe that even with successful output, the output cornmand only lasts four bit times before changing to a skip command.
Ordinarily, the F3 Hip-flop 56 serves as part of the accumulator loop with the number word appearing at its output from t1 to x32. During the output command We wish to print out only the low order tetrad of the accumulator, which ordinarily is available in the F3 flip-iiop 56 at t1 to t4. For efficient minimization, however, we wish to use the F3 ip-op part of the tetrad counter for output selection. Thus we tap the accumulator loop one bit early at the L2 flip-flop 52, where the desired information will appear at I.) to t3 rather than t1 to tt. Using the combination of F2 and F3 to count to to r3 has already been described in connection with Table 2 above.
After the output command changes into a skip cornrnand, thc flip-Hop resumes its normal role as part of the A41 rccirculating loop. Thus the remainder of the accumiuator is rccencrated and the low order terad is printed and clear. However. one transition bit must be sct up correctly when the normal regeneration resumes. During the t to f3, the net L1 [FIGURE 6N] will produce Os since no logic is written there which responds to 'the command refr'ster nal combination reprcscnting the output instruction. Then from t4 to fw the tunnel net will take the output of F3 because of the skip com d. From t5 to In, F3 truly represents the remainder ofthe information from the accumulator (passing into F3 from L2). At tit. however, F3 represents an inherited liif. from the last state. t3, of the tetrad count. Since it is indicated to cicar both the low order tetrad and the ting bit, it is necessary for the net Lt to produce G's from tanti. In order that the t4 bit be zcrocd, the tetrad count for I3 should include F3' rather than F3, and the tetrad counter therefore corresponds to the Gray or reflected binary code count.
The four paths for the output information from L2 to the output device during t@ to 13 are as shown in FlG- 'RE 6p as follows: LE-GtaS-Alt-GSTVEL LZ-GiSS-At- G84-vliz, LZ-G-Ai-GiriS-Ed, and vL2-G6S-ft4-Git5- Notice that the decoding for signals A28-A23 found on and gates GSE-G36 respectively, actually represent the count combinations corresponding to the four bits [o i0 f3.
The path F-G'Y-Eln represents the fifth Wire in the five wire output code yielding thirty-two combinations. A more detailed description of this fifth Wire in the output code was previousiy described in conjunction with the primary fun-:tions of the F9 tiipdiop 7S.
in order to rcali7e the successful output command from the engineering standpoint. we need a common signal which will be activated whenc 'er any successful output occurs, r: .ss of thc tive wire output code. This common print" command signal is represented by the path Gdkilti, in FIGURE 6i.
The following logical paths are provided to enable the operation of the output counte Initially F2 and F3 must be react for the rst count, t0, of F3'. Since the reset is simply the nrime (De Morgan) of the set net for both ot' these tina-Flops, we simply do not Write any logic in the set net operable at time tag of F8. At the end of tn however. the F3 Hip-llop is set. The path for this is F2-G9AGi3-G17-F3SF3 [FIGURE obi. At the end of" t1 we must set F2. This is accomplished by the path (Gi. GIS?. G3)-Gfi-G'-l:2l--SF2 [FIG- URE. Gul. The F3 diode in Gi, and the F6 diodes in Gitti). G2. and G3 make this path possible. At the end of t;| We must reset F3. Once again this is achieved by writing no logic to cover the desired time (i2) in the set net, l: should be noted that F29 and F39 represent the set nots for F2 and F3 dip-ilops respectively, and as previously stated. the structure of the logic is such that FZtl' and F33. represent the two reset nets respectively.
At the end of I2 the K (F1) lip-flop is set in order to later indicate the successful output. This path is rcpresentcd by Gi-G4t7-SF1 [FIGURE 6m). it was also necessary to reset Ft originally, at 132 of F3, in case the busy signal was present to make the output command unsuccessful. The path taken for this is (E43-G49- 10 INPUT (is) (nun) (5) he itc-y do`V 'n signal (Sib) is received, indicating a goo nput, i( is sct to l. rl contents ot the keyboard are read into the f3. r1, t2, and t3 positions ot the accumui-.i if the keyboard character is a non-numeric charttiot one ol the digits through 9), a 1 is written e highes" order bit position (32) ot the accumulator. s icmai der ot the accumulator is cleared to all D's. IC the keyboard (1o-.vn signal Sift is not received, K is reset to 0 and thc accumulator fills with either zeroes or meaningless information (depending on whether ...f'ftches Si, S2, Sal, and are open or in the act ol bouncing or being set).
ty between the two operations is that Fl (the K p) is used to later indicate whether or not the nica was succcsslul. FlGURE 4b shows a timing i .gram for both the successful and unsuccessful input. F11 is ally reset as the input operation begins. This :shed at 132 of the previous Fd by the path VFlGURE 6m] with the F3' diode making the path possible. At the end of t2, as Vl on the input timing diagram (FIGURE 4b), high it the ltey down signal SiO is received a successful input. The path is SIO-G2-G43- [FIGURE om i. Il the ltey down signal is not rcceivL Fl remains reset. indicating that the input was umuccesstul. The key down signal (515)) as received trom the keyboard is nothing more than a delayed common contact. In other words, when the lrcy down signal is received. wc are assured that the contacts providing the other input signals (S1, S2, S4, SS, and S11) have stoppebou ng and are thus sate to sample. Note that the contact for the lzey down signal Sit) itself may still n: bouncing. but this docs not matter since the system .s it only at one discrete time interval, namely I2 dicated. ..ier requirement oi the input timing system is that trop set F3 if Fl has been set at the cnd of t2. i i be set at the end of t3. The path for this FE-G-J-GtG-Slf [FIGURE (dij. Observe that of f5. both F2 and F3 acted as the tctrad as they did in the output operation. F2 and reset originally a: im ol FS. F3 was then of The paths necessary to achieve this f-ri'ricd in detail in the description of the tc ril counter ior the output operation. The tctrad counter and its associated logic are substantially similar for both the input and output operations.
the path of a bit from the keyboard to the write. head is SlWG'i'SGMGiGS/AIE [FiG. 6p|, G'SivGtES-Li' [,FIG. ont. Gld-L7 [FIG. 6d]. The lat- 'irt o? this path. namely, AEE-GM-GtSS-LV-GZU-L7 unon Alor all tive of the information bits passing from the keyboard switch contacts (Si, S2, S4, SS, and
ents the other four paths prior to t. merger at A12 in FIG. 6p. During f1, the path or" a 2 bit is SIZ-Gld-GIQLLAZ. At t2 the path of the 22 bit is Saf-*Glhlh-Glti- Y2. The path of the 23 bit during r3 is SSAGTSGESA. Finally, the path ior uic nonmuznric (not 0 through 9) indicator bit at tgz is biiGlll4Git5-A12.
The following rep It should be noticed that normally, the low order tctrad of the accumulator is applied from i( to l( to the write head. The successful input operation, however records the low order tetrad in the accumulator in the t0 lo t3 bit positions, cle-.trs the accumulator (to s) from r11-tgl, and, il necessary, inserts the non-numeric tetrad indicator bit at tag. The clearing of the accumulator from t4 to rs1 is accomplished by simply writing no logic in the net (Ll) to do otherwise.
A successful input command is usually followed by a shift left command, explained in greater detail below. This shifts the high order in bit (representing the nonnumeric tetrad indicator bit) into the K llip-ilop where the programmer can then test for its presence. The shift left command will also shift the low order tetrad from (t0 to t3), to (t1 to t4), and thus place it in its normal position in the accumulator.
The input operation normally serves as the lirst of a group of four operations. The second operation of the group will be a jump conditional, explained below, which tests Whether or not the input was successful by checking the K Hip-flop. A shift left command (as before noted) then follows the jump conditional provided the input was successful. The fourth operation of this group will be another jump conditional which will test for the pres- -t ence of the non-numeric tetrad indicator bit by means of the K flip-Hop.
SUBTRAC'I (SUB) (0110) (Il) K is reset to G initially. age is subtracted from the number in the accumulator, and the result replaces the number in the accumulator. If the general storage number is larger than the number originally in the accumulator, K is set to a 1. The number in general storage remains unchanged. The low order (to) bit of the accumulator is cleared (reset to 0). (FIGURE 3h).
The L5 gate circuit represents the sum-difference net for the adder-subtractor unit. The path of the resultant signal from L5 to the A-I write head is LS-GSS-G-Ll' [FIGURE 611], G20-L7 (FIGURE 6d].
It may be noted that in binary addition or subtraction, the surn and difference digits are identical but that the borrow and carry differ.
This is more easily noted from Table 3A below, in
Which F1 represents the carry-borrow, F2 represents the A major purpose of the K (F1) llip-ilop in our invention is to serve as the carry-borrow flip-flop for the adder subtractor unit. Consequently, at the end of an F8 executing an addition or subtraction operation, F1 will be left high (set to l) if wc have had an overflow in the addition or if we have subtracted a larger number from a smaller one. It should be observed that in general F1 does not change during an F8 word time, so that in essence it carries forward information from one F8' to the next F8. Thus if we followed a subtraction operation with a jump conditional (conditional on K1 to be explained below) we could branch the program depending on whether or not subtraction resulted in a change of sign.
The number in general storlll) As can bc seen from Table 3A above, the three inputs to the addcr-subtractor unit of FiG. 6e are F1, F2, and F3. As previously stated, F1 represents the carry-borrow ilip-llop. F2 represents the operand from general storage in the addition and subtraction operation, and its path from general storage to F2 is LIZ-(Gl, G1G9, G2, GB)HG4G7"-FZO-SF2 (FIG. (im. The F6 diode in G1, the L12 diode in Gl@ and G2, and the F4 diode in G3 allow this path to operate. Similarly, F3 represents the operand from the accumulator which is the augend and minuend in the addition and subtraction operations. The path of this operand from the A-I read head to F3 is LSLZ (FIG. 611], GS-GIZ-Gl-FSQLSFT (FIG. 611], with the FS diode in G8 allowing the path to operate.
The combining of the three inputs (Fl, F2, and F3) to form the sum (or the difference) in the addition and subtraction operations is represented by the common mixing paths (FZ-AillvF-(G'g-AZ, GSH-Ail, G81- AZZ, G32-A23) (FIG. (1p1, (Gli, G22., G23, G24)- GZS-LS (FIG. 66], the third input Fl enters into the nixing path through gates GDL-G25 as follows: F1- G1-G22, G23, GIM-G2545. The purpose of this mixing path in the adder-subtractor is simply to give a sum (or difference) on the L5 output line whenever any one or all three of the inputs (FI, FZ, and F3) are present. Thus, the sum-difference logic for L5 may be Written as set forth in Table 3B. This logical statement L5=1 can be expressed in the conventional terminology in which iand" groups are connected by dots and in which of the sum-difcrence gate at t0 is condition is Fl.F2'.F3'. Thus the low order bit of the accumulator is reset (cleared to 0) in both thc addition and subtraction operations. F2 and F3 are low at tu simply because no logic was written to set them at x32 ot` the previous F8. The carry-borrow flip-flop F1 was reset to 0 at the end of F32 of the previous FS by means of the path G48-G49-G52-ZF1 (FIG. 6m] with the F3' diode allowing the path to opcrate.
It is pertinent to note that up until this point a cornmon description (with the corresponding logic and gating paths) has served for both the addition and subtraction operation. The only difference between the two operations is the logic for setting and resetting the carryborrow iiip-op F1 (K) during F8'. It has already been observed that F1 is reset initially in both operations.
With reference to FIG. 6m, in subtraction, the path for setting F1 high as a borrow corresponding to Table 3B above is (F2, F3)-G82-A23 (FIG. 6Fl, G44G47 SFI (FIG. 6111]. The corresponding reset path for F1 in subtraction is (FZ, F3)-Gl}-A2I (FIG. 63], G5l- The initial output 0 because the input GSZ-ZFl (FIG. 6rn]. In thc addition operation the path for setting F1 high as a carry is (FZ, F3)-Gtl-A22` (FIG. 611], G42.-G43-G4i-SF1 IFIG. 6m], with the F6 diode in Gl?. allowing the path to operate. The corresponding reset path for F1 in addition is (FZ', F3'l-G79- All) (FIG. Gpl, G58-G52ZFI (FIG. 6111].
ADD (Aint) (m11) (7) K is reset to 0 initially. The number in general storage is added to the number in the accumulator, and the result replaces the number in the accumulator. If the result in the accumulator overflows (exceeds thirty-two bits) K is set to a l. The number in general storage remains unchanged. The low order bit 0f l HCCUmUllO 1S cleared (reset to 0).
A complete description of the ad dition OPCTHOH Wim the corresponding logic and gating Pftlls WHS COVcrCd 21 under the previous description c. the substraction operation. As previously mentioned. a common description serves both oi these operations aside from the logic oiC the carry-borrow flip-dop F1 (K) which was fully set forth above.
JUMP (J) (1000) (8) K. general storage. and accumulator remain unchanged. A new instruction word for general storage is brought into the instruction portion of the A-I channel during the next word time. The first (high order) tetrad of this new instruction word is executed three word times later (or equivalently four word times after the jump command was initially executed).
The jump command represents the first of tour (jump, change track, jump condition, and jump delay) displacement addressing instructions. They are control instructions (in contrast to arithmetic instructions such as addition) in that they operate on the instruction portion rather than the accumulator portion of the A-l channel. The normal routine of setting up (precessing in) a new command during and then executing it during the next word time F3 can only be applied to the arithmetic type instructions and the skip command. This results from the fact that only the accumulator is available during Fti'. The control type instructions are set up during F8 as usual, but we find it impossible to execute them the next succeeding F3 word time, since the instruction portion of the channel is simply not available during F8.
The purpose of the jump command is to transfer a new instruction word from general storage into the A-I channel. Supposedly we have already executed all of the eight commands found in the previous instruction word, and now must bring in a new instruction word in order to keep the computer sequencing. Ordinarily we would have to program the jump as the last (eighth or low order tetrad) command in the old instruction word in order to bring in a new instruction word. Thus every instruction Word would only have seven useful commands, with the eighth command being used as a jump in order to replace the instruction word with a new orteA ln our invention, however. all eight commands in the instruction word are useful. The jump command need never be programmed as the eighth command in order to transfer in a new instruction word. The reason for this is the automatic jump feature.
To the computer the automatic jump looks exactly as if a jump command had been programmed as a fietitious ninth command on the low order end of the instruction word. Though the automatic jump need not he programmed, the regular time ailotted for the execution of a jump command must also he allowed for the automatic jump when writing a program. ln other words, the programmer knows that the computer is going to introduce a ninth command, namely, the automatic jump, after all the instructions of a word have been executed. Thus we see that this computer has automatic sequential control for instruction words.
Though it is not programmed, the realization of the automatic jump feature in the computer is as follows:
When the new instruction word is transferred from general storage to the AHI channel (as a result of a jump type command), it occupies the position (L32 in the orders channel. During this transfer, however, a ls bit is always deposited into the low order (iiag) t0 position of the instruction portion of the A-l channel. The l`s bit will precess up the loop with the instruction word. It is this l`s bit that ultimately is executed as the automatic jump command.
At this point it is pertinent to observe some properties ot the instruction word precesslon in the A-I channel. This is a high order or long type precession in that the circulating information is displaced to a position four hits later in the word time after each prccession. Thus the tetrad of information representing the eighth command in the instruction word will move from i114 to 15gg after the initial prcecssion.
This preccssion is achieved during F8 by passing the A-I channel read output through the command shift register (four bit delay) and back to the A-Iw write head. This is represented by FIG. 3b and the path L8 SLE lllG. 61:1, (G19.y GlU-AGM-GW-F-SFS lFiG. 6bl, (GS4-G64-SF4)(G65G66-'7F4) IFIG. 60], (G26- G27-SFS)(G2E-G2`=-ZF5) |FEG. 6i). (G36-SF6) (G37-G3-ZF6) lFlG. dit. (G3t'!SF7)(G3'i-G5ZE7) (PEG. 6U, G62-G63-Lll' [FIG 611], G-Z-L? IFlG. 6d?. The F2' diode in G10, and the tpgz' diode in Gll. and the Al diodes in G65, G27, G25, G35, G37, G30. G34, and G62 make this path possible. The F4 diode in gate G64 is redundant logically, but is necessary for signal loading considerations. The A1 signal in this path represents the "advance" signal for thc command shift register (F4-F7). The logic for this advance signal is represented by the path G93-G94-A1 (FIG. 6121.
The path for the insertion of the l`s bit (representing the automatic jump) into the orders channel to t (when the instruction Word is being transferred from general storage to the order channel) is G74-A13 (FIG. Gpl, Gl5-Gl7-F3-SF3 (FIG. 6M, G81-A22 (FIG. 611], G56-G63-L1' IfFIG. Gull, G20-L7 (FIG. 6d'l. Note that while the ls bit is originally inserted (G15) at 132 of the previous F8', it picks up a unit delay in going through the F3 flipdlop so that it appears at the A-Iw write head (L7) at to of the following F8.
After each precess time circulation, the high order terad of the orders channel (which occupied the position rggg) is left in the command shift register, while the second highest order tetrad (tggga) is precessed up into the hielt order tetrad position. This tetrad which pre cesses oit the high order end of the A-l channel and left in the command shift register is the command that will be executed the next F3 word time.
During the precession, however, four new bits must have shifted into the low orde" end of the orders channel during m3 to displace the ini rmation toward the higher order end. As indicated by the logical gating path, we see that F7 is the path source for these bits. Thus we observe that ordinarily the old command code (which was just decoced from the command shilt register and executed during the F3 word time previous to the precession) would be the source of the tour additional bits in question. Allowing this situation to occur would continually recirculate the instruction word and require that a jump instruction he included in every instruction word. Therefore, the A?. function net (which represents the common reset signal for the command shift register) resets the commend shift register to all G`s at the ends of tgt; or" every Fti'. This is achieved in the A2 net by the path AES-G92A2 snown in FIG. 6p. The four paths from A?. to the command shift register are AZ-G66-ZF4 (FIG. 60j, iXi-Gin UWG. tijl, AZ-GSS-ZF (FIG. (til. and .it2-Ga`lf-G3,3G35Z `7 (FIG. 61'1. In the last ot the four paths, G3i inhibits the resetting of F7 during the change track and jump delay commands. The reason for this will be covered later during the description ot' these commands.
Thus wc observe that as a result of the A2 signal at tgz of the previous FS', four 0`s are shifted into the A-I channel at the low order end, umg, during every F8. Thus as the precessions of the instruction word continue, the instruction word eventually fills up with (Vs from the low order end. Another way of looking at this is that for every command (four bits) processed ott the high order end oi the instruction word into the command shift register, there is an equivalent four bits (all O's) shifted into the low order end of the word.
New let us return to the is bit that was placed in the A-I channel at the low order end t0 when the new instruction was transferred in from general storage.
Claims (1)
1. IN A DIGITAL COMPUTER, THE COMBINATION COMPRISING: A REGISTER FOR STORING BIVALUED SIGNALS ARRANGED IN A PLURALITY OF ALTERNATE GROUPS, THE FIRST GROUP OF SAID PLURALITY REPRESENTING AN ACCUMULATOR NUMBER AND THE SECOND GROUP OF SAID PLURALITY REPRESENTING INSTRUCTIONS, SAID SECOND GROUP BEING FURTHER ARRANGED IN A PLURALITY OF SUB GROUPS OF SIGNALS, AT LEAST ONE SUB GROUP OF SAID LAST NAMED PLURALITY REPRESENTING AN INSTRUCTION CORRESPONDING TO AN OPERATION TO BE PERFORMED UPON INFORMATION IN SAID COMPUTER; MEANS FOR CIRCULATING AND SERIALLY PRESENTING SIGNALS STORED IN SAID REGISTER; OPERATING MEANS CONNECTED TO SAID CIRCULATING MEANS AND OPERABLE IN RESPONSE TO A FIRST SUB GROUP OF SIGNALS REPRESENTING AN INSTRUCTION TO PERFORM A CORRESPONDING OPERATION UPON INFORMATION IN SAID COMPUTER AND, IN EACH SUBSEQUENT CYCLE OF OPERATION, BEING OPERABLE IN RESPONSE TO ANOTHER SIGNAL OF SAID SUB GROUPS FOR PERFORMING AN OPERATION CORRESPONDING THERETO, EACH CYCLE REQUIRING ONLY THE TIME NECESSARY FOR READING THE SIGNALS OF ONE EACH OF SAID FIRST AND SECOND GROUPS.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US856183A US3116410A (en) | 1959-11-30 | 1959-11-30 | Simple general purpose digital computer |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US856183A US3116410A (en) | 1959-11-30 | 1959-11-30 | Simple general purpose digital computer |
Publications (1)
Publication Number | Publication Date |
---|---|
US3116410A true US3116410A (en) | 1963-12-31 |
Family
ID=25323019
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US856183A Expired - Lifetime US3116410A (en) | 1959-11-30 | 1959-11-30 | Simple general purpose digital computer |
Country Status (1)
Country | Link |
---|---|
US (1) | US3116410A (en) |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US3327294A (en) * | 1964-03-09 | 1967-06-20 | Gen Precision Inc | Flag storage system |
US3331954A (en) * | 1964-08-28 | 1967-07-18 | Gen Precision Inc | Computer performing serial arithmetic operations having a parallel-type static memory |
US3348215A (en) * | 1961-12-27 | 1967-10-17 | Scm Corp | Magnetic drum memory and computer |
US3404375A (en) * | 1964-04-02 | 1968-10-01 | Hughes Aircraft Co | Combination random access and mass store memory |
US3473161A (en) * | 1966-11-23 | 1969-10-14 | Gen Electric | Circular listing |
US3763475A (en) * | 1972-04-12 | 1973-10-02 | Tallymate Corp | Stored program computer with plural shift register storage |
US3775754A (en) * | 1968-04-10 | 1973-11-27 | H Auspurg | Dial-operated data exchange system |
US3987419A (en) * | 1974-12-05 | 1976-10-19 | Goodyear Aerospace Corporation | High speed information processing system |
US4012721A (en) * | 1975-05-23 | 1977-03-15 | General Electric Company | Digital logic circuit for a dynamic buffer register |
US8042284B2 (en) * | 2006-10-09 | 2011-10-25 | Lg Electronics Inc. | Heating system, drying machine having the heating system, and method of controlling the heating system |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US2954166A (en) * | 1952-12-10 | 1960-09-27 | Ncr Co | General purpose computer |
US2978175A (en) * | 1953-02-11 | 1961-04-04 | Ibm | Program control system for electronic digital computers |
-
1959
- 1959-11-30 US US856183A patent/US3116410A/en not_active Expired - Lifetime
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US2954166A (en) * | 1952-12-10 | 1960-09-27 | Ncr Co | General purpose computer |
US2978175A (en) * | 1953-02-11 | 1961-04-04 | Ibm | Program control system for electronic digital computers |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US3348215A (en) * | 1961-12-27 | 1967-10-17 | Scm Corp | Magnetic drum memory and computer |
US3327294A (en) * | 1964-03-09 | 1967-06-20 | Gen Precision Inc | Flag storage system |
US3404375A (en) * | 1964-04-02 | 1968-10-01 | Hughes Aircraft Co | Combination random access and mass store memory |
US3331954A (en) * | 1964-08-28 | 1967-07-18 | Gen Precision Inc | Computer performing serial arithmetic operations having a parallel-type static memory |
US3473161A (en) * | 1966-11-23 | 1969-10-14 | Gen Electric | Circular listing |
US3775754A (en) * | 1968-04-10 | 1973-11-27 | H Auspurg | Dial-operated data exchange system |
US3763475A (en) * | 1972-04-12 | 1973-10-02 | Tallymate Corp | Stored program computer with plural shift register storage |
US3987419A (en) * | 1974-12-05 | 1976-10-19 | Goodyear Aerospace Corporation | High speed information processing system |
US4012721A (en) * | 1975-05-23 | 1977-03-15 | General Electric Company | Digital logic circuit for a dynamic buffer register |
US8042284B2 (en) * | 2006-10-09 | 2011-10-25 | Lg Electronics Inc. | Heating system, drying machine having the heating system, and method of controlling the heating system |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US2916210A (en) | Apparatus for selectively modifying program information | |
US4079451A (en) | Word, byte and bit indexed addressing in a data processing system | |
US3593313A (en) | Calculator apparatus | |
US4135242A (en) | Method and processor having bit-addressable scratch pad memory | |
US3328768A (en) | Storage protection systems | |
US4631663A (en) | Macroinstruction execution in a microprogram-controlled processor | |
US3585605A (en) | Associative memory data processor | |
US3161763A (en) | Electronic digital computer with word field selection | |
GB1274830A (en) | Data processing system | |
US3215987A (en) | Electronic data processing | |
US4047247A (en) | Address formation in a microprogrammed data processing system | |
CA1093214A (en) | Microprogram address dualing | |
US4079447A (en) | Stored program electronic computer | |
JPS6351287B2 (en) | ||
JPS598846B2 (en) | Microprogrammable peripheral controller | |
US3116410A (en) | Simple general purpose digital computer | |
US3228005A (en) | Apparatus for manipulating data on a byte basis | |
US3015441A (en) | Indexing system for calculators | |
US3341817A (en) | Memory transfer apparatus | |
GB1003923A (en) | Digital computing system | |
US3213427A (en) | Tracing mode | |
US3263218A (en) | Selective lockout of computer memory | |
US3840864A (en) | Multiple memory unit controller | |
EP0649083A2 (en) | A microcontrol unit for a superpipelined, superscalar microprocessor | |
GB1003921A (en) | Computer cycling and control system |